Original Research

The development and validation of two equivalent organisational leadership scales from an Afrocentric perspective

Anton Grobler, Eben Enslin

Received: 04 Sept. 2025; Accepted: 04 Nov. 2025; Published: 13 Jan. 2026

Copyright: © 2026. The Author(s). Licensee: AOSIS.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

Orientation: Organisational leadership (OL) is a vital yet complex, context-specific phenomenon, mainly conceptualised and measured from a Western-centric perspective.

Research purpose: This study develops and validates a context-specific instrument, the organisational leadership scale (OLS), to measure OL.

Motivation for the study: Leadership is often treated as a universal concept, measured by decontextualised Western instruments, overlooking contextual nuances. This study addresses this gap by creating an emic tool for OL assessment.

Research approach/design and method: Adopting a positivistic paradigm, this quantitative, cross-sectional research comprises two studies. Study 1 used interactive qualitative analysis (IQA) to identify 32 unidirectional items. Study 2 validated a refined 19-item instrument, testing both unidirectional and multidirectional items, with measurement invariance assessed across private and public sectors.

Main findings: Two equivalent OLS versions were developed: a unidimensional model with 19 unidirectional items and a bifactor model with constructive and destructive OL sub-factors (multidirectional items). Wording direction slightly influenced responses, but no measurement invariance was found.

Practical/managerial implications: The OLS offers reliable tools for leadership development and coaching, enhancing methodological approaches through multidirectional items and bifactor analysis.

Contribution/value-add: This study provides a context-specific, emic instrument for assessing OL, available in two validated versions, advancing leadership research and practice, also globally, for instance, in similar collectivistic cultures.

Keywords: organisational leadership; Afrocentric leadership; authentic leadership; transformational leadership; leadership scale; leadership measurement.

Introduction

Organisational leadership (OL) is a critical driver of organisational success, shaping performance, culture and employee engagement in dynamic and context-specific environments (Northouse, 2021; Yukl & Gardner, 2020). Despite extensive global research, the application of leadership theories in African contexts, also in South Africa, remains underexplored, with a heavy reliance on Western-centric frameworks that often fail to capture the unique socio-cultural, economic and political dynamics (Banutu-Gomez Fanneh et al., 2023; Enslin & Grobler, 2025). These universal models risk oversimplifying leadership by neglecting local values and practices, necessitating context-specific (emic) approaches that integrate African perspectives, such as Ubuntu, to accommodate the cultural diversity (Abdulai, 2021; Grobler & Singh, 2018; Mbandlwa & Fagbadebo, 2020), thus acknowledging the uniqueness of the context when studying leadership. This study’s specific context, unlike developed approaches that prioritise tasks, promotes ethical and inclusive practices in organisations, resonating with collectivist and relational leadership styles and offering improved employee engagement along with broader global relevance.

According to Creighton-Randall (2024) and Vika and Pule (2025), organisations across the globe are increasingly valuing environments built on mutual respect, empathy and shared responsibility to boost performance and strengthen social bonds. As global challenges grow more intricate, human-centred leadership provides a powerful path to ethical, inclusive and effective practices across the world. The plea for context-specific leadership research extends beyond Africa; Johns (2024) advocates for context-specific leadership and urges a departure from blindly relying on universal or decontextualised frameworks. This shift is crucial for both academic theory and practical leadership development (Cikara et al., 2022; Johns, 2017), and, related to this study, the accurate measurement of leadership (Grobler & Sibanda, 2024; Grobler & Singh, 2018).

This study is focused on the development and validation of a scale that measures OL, and is conceptually based on (1) a thorough systematic literature review (Enslin & Grobler, 2025) and (2) a contextual (emic) investigation into OL, using interactive qualitative analyses (IQA) (Grobler & Enslin, in press). The IQA was purposefully used by Grobler and Enslin (in press) as a technique that is based on participation and collaboration, intended to explore local values, worldviews and lived experiences of organisational leaders as well as ‘the consumers’ of leadership. Their findings reflect the contextual realities of OL, supported by various scholars from the African continent, for example, Iyioke (2017), Punnett (2017), as well as Vilakati and Schurink (2021).

Discussion of organisational leadership from an Afrocentric perspective

Northouse (2021) and Yukl and Gardner (2020) define OL as the process of influencing and enabling individuals or groups to achieve shared organisational goals, a dynamic and context-dependent phenomenon. The evolution of leadership theories and approaches, from trait and behavioural to contingency, for instance, transformational and servant leadership, has enriched global understanding of the phenomenon, but requires adaptation to African contexts (Campos, 2024; Eva et al., 2019).

In the African contexts, leaders navigate multifaceted challenges, including socio-economic disparities, historical inequalities and cultural diversity, which demand adaptive and transformative leadership approaches (Mbandlwa & Fagbadebo, 2020). However, some leadership theories, such as transformational leadership, which emphasises inspiration and development, and servant leadership, with its focus on empathy and community (Avolio & Bass, 2004), are in some way aligned with African values like Ubuntu (Grobler & Sibanda, 2024). However, the over-reliance on foreign (etic) frameworks and the resulting scarcity of culturally specific instruments hinder progress in African leadership research (Enslin & Grobler, 2025). For instance, only 12% of 131 African leadership studies reviewed by Enslin and Grobler (2025) utilised emic methodologies, underscoring the need for contextual research and development of Afrocentric leadership in general and, more specifically, related to this study, OL.

Afrocentric leadership centres around African values, interests and perspectives, promoting a community-focused approach that prioritises fairness, harmony, participation and collective decision-making (Asante, 1988; Grobler & Singh, 2018; Laloo, 2022; Mangaliso et al., 2021; Mbigi, 2007). Similarly, Ubuntu (as a specific expression of Afrocentric leadership) is a philosophy emphasising personhood through relationships, rooted in principles of morality, interdependence, respect and inclusivity (Mbigi, 2007; Mbiti, 1989). In contrast to Western leadership models, which often highlight individualism and competition, Afrocentric leadership fosters teamwork and collective kinship, aligning with globally acknowledged leadership frameworks, such as servant and authentic leadership (Chawane, 2016; Geldenhuys & Veldsman, 2011). These values manifest in organisational settings, where leaders act as servants for the collective good, enhancing social cohesion and organisational effectiveness (Zondo, 2022).

This study acknowledges context through appropriate instrument development and validation by using the contextual results obtained from an IQA process (Enslin, 2023; Grobler & Enslin, 2021, 2022). This contribution counters ethnocentrism by engaging participants in defining OL based on their lived experiences, revealing key elements of the phenomenon and their causal relationships through system influence diagrams (Grobler & Enslin, in press). This study differs from leadership studies that followed a mainly decontextualised quantitative research approach (Enslin & Grobler, 2025). Decontextualised leadership studies contribute to the scarcity of emic (context-specific) leadership scales and over-reliance on Western instruments, which is generally regarded to limit the validity of findings in diverse settings (Meiring et al., 2005). The reality is that universal leadership instruments often fail to capture culturally embedded meanings. This challenge should be addressed through the advancement of emic-related (context-specific) research or at least a balanced emic-etic approach that integrates contextual dimensions across contexts (Grobler & Grobler, 2023, 2024; Walumbwa et al., 2008). This emic-etic integration is based on the relational leadership framework encompassing universally acknowledged styles like authentic, transformational and servant leadership. It views leadership as a collaborative social process shaped by interactions, emphasising emergent coordination, open communication and multidirectional influence that transcends conventional leader-follower hierarchies (Tabassi et al., 2025).

A total disregard for contextualised research and the over-reliance on decontextualised quantitative methodologies (instruments), however, result in the contextual nuances often being neglected. In this study, the IQA is proposed as a participatory, structured method for developing contextually relevant frameworks (Enslin & Grobler, 2025), in this case, OL within the South African (SA) context. The SA context is characterised by a diverse ethnic, linguistic and socio-economic landscape that demands leadership models that reflect local realities (Mbandlwa & Fagbadebo, 2020). The internal realities are further exacerbated by global challenges like digital transformation and volatile, uncertain, complex and ambiguous (VUCA) environments, requiring leadership that is culturally aware and adaptable to the challenges (Johns, 2024; Tabrizi et al., 2019).

Research purpose

In support of context-specific leadership research in SA, this study aims to develop and validate two equivalent versions of an emic organisational leadership scale (OLS), across two independent studies. One version of the OLS has only positive worded items (unidirectional), and the other consists of the same items, but worded multidirectionally (positive and negative).

The objectives of the overall study are, firstly, the development and validation of an OLS that consists of positively worded items (Study 1); secondly, to assess measurement invariance across sectors (private and public sectors); thirdly, the development and validation of an OLS instrument, with multidirectional items to address possible acquiescence and common method bias associated with unidirectional items and to improve measurement balance (Study 2); and lastly, to assess the impact of the difference between the two versions of the OLS instrument (Study 1&2).

Potential value added by the study

The unique contribution of this study is the development and validation of two equivalent scales that measure OL, developed from an emic perspective, thus being context-specific. This study’s practical contribution is the provisioning of a scientifically validated OLS, offered in two equivalent versions: one comprising unidimensional items and the other incorporating multidirectional items. On a theoretical level, the study contributes to a more inclusive and contextually grounded body of leadership theory, offering conceptual insights into OL. Methodologically, the application of bifactor analysis alongside traditional exploratory and confirmatory factor analyses (EFA and CFA), as well as the comparison between the two versions, represents a novel approach in leadership research and may serve as a valuable reference for future scholarly investigations into the leadership construct.

Research design

Research approach

This study employed a typical positivistic, empirical paradigm using a cross-sectional design, surveys and quantitative analysis.

Research participants

The study population consists of SA employees working in organisations with 60 or more staff members. Participants were purposively sampled, with 60 respondents selected from each participating organisation. The sample characteristics are presented in Table 1.

The total sample consisted of 4192 respondents, with Study 1 accounting for 60.9% (n = 2554) and Study 2 for 39.1% (n = 1638), drawn from 70 organisations. Both private and public sector organisations participated, with the public sector better represented (38 organisations) compared to the private sector (32 organisations), comprising 53.9% and 46.1% of respondents, respectively. The combined sample had a mean age of 38.09 years (standard deviation [SD] = 9.97) and an average organisational tenure of 8.34 years (SD = 8.68), with slight variations in age and tenure across the studies. Males made up 50.7% of the sample. Most respondents were in non-managerial roles (64.2%) and support functions (56.6%). The racial composition mirrored the general SA workforce, with 72.3% from the African race category. Nearly 96% of respondents across the combined sample had at least a Grade 12 education, suggesting proficiency in English despite it not being the first language for most. Overall, the respondents appear mature, experienced and educated, qualities essential for providing informed perspectives on leadership perceptions within their organisations.

Research procedure

A two-phase exploratory design was used to develop the OLS. Firstly, a thorough systematic literature review was conducted by Enslin and Grobler (2025). The rationale for this step was to establish how OL is conceptualised and defined and how it has been measured up to now. This was followed by an IQA process that resulted in a conceptualisation of OL from an emic perspective (Grobler & Enslin, in press). They reported that OL comprises a primary driver, namely the leader’s emotional awareness; secondary drivers, namely the leadership style, characteristics, culture, communication and vision. They further reported that these drivers resulted in leader support and team dynamics (as a secondary outcome) and delivering strategy as a primary outcome. The elements of affinities were well defined by the participants in the IQA process, and based on that, the 32 (positive worded) items were developed, based on the qualitative information from the IQA process, to measure OL.

Data were collected by means of the OLS (and related instruments that are discussed later). The 32 unidirectional (positive) worded items were initially used in Study 1, and the refined version consisting of 19 multidirectional items was used in Study 2. This was done by converting the existing six items into negative-worded items. The essence of the items remained the same compared to those used in Study 1, just reverse-keyed. The rationale for the adaptation of the OLS instrument in Study 2 was to provide an OLS version that could mitigate acquiescence bias (Bulut & Bulut, 2022), enhance content validity, reduce the risk of common method bias (Podsakoff et al., 2003) and improve measurement balance (Kam & Meyer, 2015). Bulut and Bulut (2022) warn, however, against factor structure distortion, lower reliability and internal consistency and reduced criterion validity when the direction of the items is varied.

Statistical analysis

The statistical analysis was performed by using the Statistical Package for the Social Sciences (SPSS 29), supported by SPSS Amos (Analysis of Moment Structures, version 25) (IBM, 2023). The Bifactor Indices Calculator (Dueber, 2017) was used with the bifactor model evaluation.

Data screening

The dataset was first cleaned up by means of case screening, followed by variable screening to ensure that there were no missing values in the dataset. The dataset was further inspected for unengaged responses through an inspection of the SD and to delete cases with SD < 0.50. From the data cleaning process, it was deduced that the missing values were very sparse (less than 5% of the cases were deleted) and they were, therefore, not considered a main contributor to any bias. The variables were further screened by means of kurtosis and skewness with critical thresholds of 2.00 and 7.00, respectively, to be considered normally distributed (West et al., 1995).

Exploratory factor analysis

The first step of the factor analysis was to evaluate the appropriateness of the sample size to conduct an Exploratory factor analysis (EFA). The item to respondent ratio of ±1:20 is regarded as acceptable according to Meyers et al. (2013). Secondly, the intercorrelations between items were inspected using Bartlett’s test of sphericity (Hair et al., 2019). With this test, the statistic generated should be significant (p < 0.05) for an EFA to be considered an appropriate technique (Hair et al., 2019). Thirdly, the Kaiser–Meyer–Olkin (KMO) measure was used to quantify whether the items correlated sufficiently to determine whether a factor analysis could be performed. The minimum level set for this statistic is 0.60 (Tabachnick et al., 2013).

To aid in the interpretation of the initial results, orthogonal rotation, specifically the Promax rotation, was used. The decision regarding the number of variables (factors) to be retained was based on the Guttman–Kaiser eigenvalue greater-than-one rule (K1 rule), together with the scree plot (with specific reference to the shape of the curve) and, lastly, the Monte Carlo principal component analysis (PCA) for parallel analysis. Meyers et al. (2013) indicate that a guide for variance accounted for by the factors needs to meet the lower limit of 50%. The Cronbach alpha coefficient was determined taking into consideration that the general rule according to Nunnally and Bernstein (1994) is α > 0.70.

Confirmatory factor analysis

To operationalise this construct definition of the construct, namely OL, a CFA was conducted. To assess the model fit, several fit indexes were used, including the comparative fit index (CFI), the root mean square error of approximation (RMSEA), Chi-square (χ²) and the ratio of the differences in Chi-square to the differences in degrees of freedom (χ²/df). Given that there is no one acceptable cut-off value for what constitutes adequate fit, it was elected to evaluate the model and to recommend the model. The CFA index values recommended by Byrne (2016) are 0.90 for the CFI value, an RMSEA value of 0.05 and, in terms of the χ²/df, a ratio of less than 5.00. The option to use only these indices is supported by Cheung and Rensvold (2002), who regard it as a suitable indication of good fit.

Validity assessment

Convergent validity of the items was assessed by the composite reliability (CR) and the average variance extracted (AVE), with critical values of >0.70 and >0.50, respectively. The discriminant validity was determined by comparing the AVE with the maximum shared variance (MSV). Proof of discriminant validity would be apparent when MSV < AVE and where the average shared variance (ASV) is less than AVE (Hair et al., 2019).

Measurement invariance at the sectoral level was evaluated with regard to the unidirectional instrument in Study 1. It was conducted across the private and public sectors, focusing on three levels: configural, metric and scalar invariance. Configural invariance tests whether the same underlying factor structure as measured by the instrument represents OL similarly across the sectors. Metric invariance (also referred to as weak invariance) extends this by requiring that the factor loadings are equivalent across sectors, indicating that the strength of the relationship between items and the latent construct is consistent. For the purpose of this study, changes in fit indices, rather than traditional Chi-square difference tests, were used to assess invariance because of the Chi-square statistic’s sensitivity to large sample sizes. Scalar invariance (or strong invariance) further requires that item intercepts are equal across sectors, implying that not only the meaning (factor loadings) but also the levels (intercepts) of the items are comparable between groups. Establishing scalar invariance allows for meaningful comparison of latent means across sectors. Following the recommendations of Cheung and Rensvold (2002), a change in CFI of no more than 0.01 was used as the threshold for invariance. In addition, Chen (2007) advises the use of alternative fit indices, specifically changes in RMSEA (ΔRMSEA ≤ 0.015) and SRMR (ΔSRMR ≤ 0.030), to evaluate metric and scalar invariance.

To examine the construct’s dimensionality in Study 2, using the OLS consisting of multidirectional items, a bifactor model was applied, utilising various indices to assess the strength of general versus group factors (Rodriguez et al., 2016). These included omega (ω) for total reliability; omega hierarchical (ωH) for general factor variance; omega subscale (ωS) and omega hierarchical subscale (ωHS) for sub-factor reliability; factor determinacy (FD) to gauge how well scores reflect true latent variables; and construct replicability (H), which reflects the consistency of factor representation.

Explained common variance (ECV) further informs the level of unidimensionality. A general factor is favoured when ωH > 0.80, ECV > 0.70, PUC > 0.80, and the difference between ω and ωH is minimal (Stucky & Edelen, 2014). Factor determinacy values > 0.90 suggest reliable factor score estimates, although both FD and H can be inflated by a small number of highly loading items (Gorsuch, 1983). Together, these indices help determine whether a unidimensional or multidimensional model best captures the construct’s structure.

Assessment of the difference between the two versions was done statistically on the item level, and correlations of OL (and in Study 2, the factors) with cognate instruments and variables. Correlations between the constructs were calculated by means of Pearson’s product-moment correlations. To assess the statistical significance of the difference in the correlations across sectors, z-observed values were calculated, following a Fisher r-to-z transformation. Interpretation was based on the z-observed value, with the difference in the correlations deemed to be statistically significant when greater than 1.96 or smaller than −1.96. The instruments used for this step in the analysis are the adapted authsentic leadership instrument (Grobler & Grobler, 2024; Walumbwa et al., 2008), ethical leadership (De Hoogh & Den Hartog, 2008; Grobler, 2017), servant leadership (Grobler & Flotman, 2021; Liden et al., 2015), spiritual leadership (Grobler & Sibanda, 2024) and transformational leadership (Grobler & Grobler, 2023; Podsakoff et al., 1996).

The same correlations, as used to determine the difference in the relationship between the constructs measured by the respective versions of the OLS, were used as a determination of convergent validity. A correlation of 0.40 is an indication of convergence, with 0.50 and higher a clear sign of convergence (Cohen et al., 2013) and is often referred to as heterotrait-monomethod coefficients (HTMM coefficients). This entails correlations between measures of different traits that are furnished by the same method of measurement.

Ethical considerations

Ethical clearance was obtained from the institution’s research ethics committee, which includes permission from each of the organisations and consent from all the participants. This research did not receive any specific grant from funding agencies in the public, commercial or non-profit sectors. The ethics approvals are: 2020_SBL_AC_009_CA, dated 01 June 2020 and 2022_SBL_AC_004_CA, dated 20 May 2022.

Results

The results of the statistical procedures discussed in the previous section are now reported in this section. To determine the construct validity and construct equivalence of the OLS, an EFA was conducted. The EFA was first conducted on 32 items of the OLS with all being positively worded in Study 1 and then on a combination of 19 positive and negative worded items in Study 2.

The relatively large sample size contributed to an adequate variable-to-respondent ratio (80:1 and 86:1 in Study 1 and Study 2, respectively). It must be noted that Study 1 examined the 32-item OLS instrument, while Study 2 examined the refined 19-item version. The values of the KMO measure of sampling adequacy were 0.98 and 0.95, above the recommended value of 0.60, and Bartlett’s test of sphericity was significant (p < 0.001), and χ² (171) = 37515.05 and χ² (171) = 15984.85, supporting the strategy to conduct an EFA in both studies.

The Guttman–Kaiser K1 rule was used in conjunction with the scree plot to determine the number of factors. The results of the Kaiser’s criterion, as determined by the PCA, are reported in Table 2.

The factor solution with eigenvalues larger than one, reported in Table 2, consists of one factor consisting of 19 items, and two factors for Study 1 and Study 2, respectively. The total variance explained exceeds the critical value of 50%. Cattell’s scree test, which is aimed at retaining the components (factors) before the break (elbow rule), was performed, and the results are reported in Figure 1.

FIGURE 1: Cattell’s scree plot.

The interpretation of the scree plots, presented in Figure 1, suggests the presence of a single dominant factor in Study 1, as indicated by the clear inflection point (elbow) after the first component. In contrast, the scree plot for Study 2 suggests the presence of two distinct factors.

An HTMT value of 0.46 (< 0.85, Kline, 2011) was reported in Study 2, indicating that discriminant validity is established and that the two factors are distinctively different from one another. To further assess construct validity of the instruments, CFAs were conducted for both studies. The results reported in Table 3 include key model fit indices, namely the CFI, RMSEA, Chi-square (χ²) and the ratio of the change in Chi-square to the change in degrees of freedom (Δχ²/Δdf). Study 1 consists of 19 items, all positively worded, and Study 2, with the same items, except that 6 have been changed to be negatively worded, yielding a two-factor factor structure.

Consequently, a CFA was conducted to validate the internal structure of the scale derived from the EFA for both Studies^1&2. Study 1 consisted of the confirmation of the unidimensional model, while Study 2 assessed four competing models. These statistics fit these models, namely a unidimensional (one-factor) model, a first-order three-factor model, a second-order three-factor model and a bifactor model, as reported in Table 3.

The CFA in Study 1 confirmed the unidimensional factor model (all 19, unidirectional items) of the OLS (χ²/df [184] = 21.75, CFI = 0.96, RMSEA = 0.065). The analysis of Study 2 was more complicated by the inclusion of multidirectional items, as it impacted on the factor structure. A first-order CFA with a two-factor solution was conducted to validate the measurement model and yielded acceptable fit statistics (χ²/df [146] = 6.48, CFI = 0.95, RMSEA = 0.058). A comparative analysis of the model fit indices revealed that the first-order factor model exhibited superior fit compared to the unidimensional model (Chi-square difference between these two models was 1322.71 [Δχ²= 2511.69–888.98] with a corresponding difference of 1 degree of freedom [152−151]). This was followed by attempts to introduce a second-order factor, but this resulted in estimation errors, and it was therefore decided to conduct bifactor analysis, as it would assess the possibility of an overall OL factor.

The bifactor model demonstrated the best fit (χ²/df [133] = 4.81, CFI = 0.95, RMSEA = 0.058). The Chi-square difference between the first-order and bifactor models was 249.12 (Δχ²= 888.98–639.86) with a corresponding difference of 18 degrees of freedom (151−133). In this model, all 19 items were set to load concurrently on a general factor, representing overall OL, and on two orthogonal group factors, namely constructive and destructive OL, respectively. This structure enabled the separation of item variance, firstly, into shared variance from the general factor (OL) and unique variance from each specific sub-factor, providing a clearer and more detailed understanding of OL by capturing its multidimensional nature.

A further examination of OL’s dimensionality (as a bifactor model) was conducted by assessing the strength of a general versus sub-factors, using various indices. The results are reported in Table 4.

The omega (ω) estimates for the general factor, representing OL, was exceptionally high at 0.95, indicating a highly reliable multidimensional composite. This suggests that the general factor captures most of the systematic variance in unit-weighted total scale scores. Sub-factor scores, comprising destructive and constructive OL, demonstrated slightly lower reliability, with ωS ranging from 0.84 to 0.95 (mean ωS = 0.90, SD = 0.06). This indicates that sub-factor reliability is adequate but less robust than the general factor, as expected because of item count differences (Reise et al., 2013). The variance attributed to multidimensionality from sub-factors is lower, reinforcing the general factor’s dominance.

The omega hierarchical (ωH) value of 0.84, exceeding the 0.80 threshold (Reise et al., 2013), further supports the general factor that accounts for the most reliable variance. After controlling for the general factor’s variance, the omega hierarchical subscale (ωHS) for sub-factors averaged 0.33 (SD = 0.29), suggesting that much of the reliable variance in sub-factor scores derives from the general factor (OL) rather than the sub-factors. While sub-factor scores remain sufficiently reliable for basic research, their low ωHS relative to ωS warrants cautious interpretation.

The FD value for the general factor was 0.96, indicating that observed individual differences in factor scores closely reflect true differences on the factor. Sub-factor FD values were lower (0.89 and 0.82; mean = 0.86, SD = 0.04), influenced by fewer items per sub-factor (Rodriguez et al., 2016). The H values for sub-factors averaged 0.70 (SD = 0.07). The FD and H values confirm the general factor’s dominance over sub-factors, but support the utility of the sub-factors for research purposes.

Finally, to evaluate potential bias from fitting multidimensional data to a unidimensional model, the ECV was assessed. Although the ECV value of 0.66 is slightly below the 0.70 threshold (Rodriguez et al., 2016), it still supports the use of OL as a general factor.

The descriptive statistics, the internal consistency as well as the convergent and discriminant validity of OL as a unidimensional construct (Study 1) and a two-factor with a general factor (bifactor model) (Study 2) are reported in Table 5.

The unidimensional OL construct in Study 1 exhibited a relatively high mean score (M = 3.73, SD = 0.82), while the general factor showed a slightly lower mean (M = 3.60, SD = 0.69), possibly because of the inclusion of negative items in Study 2. Variables based on positively worded items (OL and constructive OL) displayed negative skewness, indicating distributions with fewer small values and a left-tailed tendency. Conversely, destructive OL showed a positive skewness value of 0.26. Kurtosis values ranged from −0.61 to 0.60. As the skewness and kurtosis values for all factors remained below the critical thresholds of 2.00 and 7.00, respectively, the data can be considered normally distributed.

Cronbach’s alpha coefficients for all variables across both studies are deemed acceptable (α ≥ 0.70), indicating satisfactory internal consistency (Hair et al., 2019). Convergent validity was established through Composite Reliability (CR) and Average Variance Extracted (AVE). Discriminant validity was assessed and confirmed by comparing AVE with MSV, with satisfactory results indicated by MSV being less than AVE and ASV also being less than AVE (Hair et al., 2019).

The subsequent step in the scale development involved examining measurement invariance, which was conducted in Study 1. Measurement invariance ensures that the OLS measures the same construct in the same way across different groups (Davidov et al., 2014), in this case, the private and public sectors. A random sample of 500 participants from each group (private and public sectors) was used to analyse invariance. The analysis employed a criterion of a change of −0.01 in the CFI, with Chen (2007) recommending the use of alternative fit indices such as ΔRMSEA and ΔSRMR. The critical values for these indices were set at 0.015 and 0.030 for metric invariance and scalar invariance, respectively.

Results supported configural invariance, as indicated by acceptable model fit statistics when the groups were estimated freely, without constraints. Specifically, the changes in fit indices were ΔCFI = 0.01, ΔRMSEA = 0.017(marginally exceeds the 0.015 criterion), and no change was observed in SRMR.

Further assessment of measurement invariance from both a metric and scalar perspective was conducted, and the detailed results are presented in Table 6.

The results reported in Table 6 indicate that metric and scalar invariance is supported, as the delta values of the fit indices between unconstrained and fully constrained models meet the recommended values. The instrument can therefore be used within the private and public sectors as it measures OL consistently in the same way across the sectors.

The fourth research objective is to compare the two versions, focusing on the impact of the directional change of six items in Study 2. These items were merely transformed from being positively to negatively worded, without changing the essence of the item. The comparison was performed on the item level (see Table 7) and on the construct level, as reported in Table 8. The items that were changed are marked with an asterisk (*).

The comparison of the two versions of the OLS indicates that the most significant differences are reported on four of the reversed items, namely items 4 (∆ = 38.1%), 14 (∆ = 20.1%), 16 (∆ = 19.6%) and 18 (∆ = 14.8%).

Only 14% of the respondents responded overall negatively on OL in Study 1 compared to the 18% in Study 2, resulting in the positive response being 4.58% less in Study 2. Presented in a different manner, the difference between the negative and positive responses is presented as a ratio of 1:4.95 and 1.3.55 for Study 1 and Study 2, respectively.

In summary, participants in Study 1 (one-directional) perceived their leaders more positively, with the largest differences emerging in reverse-coded items (Study 2) related to empathy, diversity, emotional safety and belonging.

To assess the impact of incorporating multidirectional items on the relationship between latent constructs and external criterion variables, as postulated by Podsakoff et al. (2003), a preliminary (albeit limited) analysis was conducted to examine the relationship between OL as measured by the one-directional instrument and the OL bifactor derived from the multidirectional instrument.

To explore this concern, basic correlational analyses were conducted with related leadership constructs that are theoretically expected to be associated. The results are presented in Table 8.

It is apparent that there is a degree of convergence between OL (both variations) and the cognate leadership constructs. Comparing the two versions indicates that the differences in the correlations are minor, with only the correlations between transformational leadership and the OL bifactor being significantly stronger than the correlation with OL in Study 1 (r = 0.77 and r = 0.85; z-score of −7.44). The inclusion of the multidirectional items, in this case, did not have an impact on the relationship with external variables. No evidence could be found that the change in the direction of the items lead to weaker correlations between the latened variables and external criterion variables (Podsakoff et al., 2003).

Discussion and conclusion

Organisational leadership is regarded as pivotal in driving organisational success by shaping performance, culture and employee engagement within dynamic, context-specific settings. However, global leadership research, predominantly rooted in Western frameworks, often overlooks the unique socio-cultural, economic and political dynamics of African contexts, including SA. These universal models risk oversimplifying leadership by neglecting local values such as Ubuntu, which emphasises interdependence and community. The limited application of context-specific (emic) approaches in African leadership has led to the need for frameworks that integrate African perspectives to address cultural diversity and contextual challenges effectively.

This study advances context-specific leadership research through the development and validation of an OLS – more specifically, two equivalent versions thereof. The approach to developing the instrument is regarded as emic in nature, as it was based on a systematic literature review and IQA (Enslin & Grobler, 2025; Grobler & Enslin, in press). By engaging leaders and their followers to explore local values, worldviews and lived experiences, the IQA fostered a participatory approach to understanding OL’s contextual realities. This approach is necessitated because of the uniqueness of leadership in Africa, which is grounded in principles like Ubuntu, prioritises collective decision-making, fairness and social cohesion, and aligns with servant and authentic leadership models. This contrasts with Western models’ focus on individualism and competition, emphasising the necessity of emic methodologies to develop culturally relevant leadership instruments and counter the limitations of decontextualised, universal frameworks in diverse settings.

This study focuses on developing and validating an emic OLS, tailored to the SA context, with two equivalent versions evaluated across two independent studies. The first version (Study 1) consists of only positively worded items (unidirectional), while the second (Study 2) uses the same items but worded multidirectionally (positive and negative). The rationale for this conversion was to mitigate issues like acquiescence and common method bias, while enhancing measurement balance.

The study’s objectives are fourfold. Objective 1 was to develop and validate the unidirectional item OLS. An instrument, consisting of 32 original items, was assessed in terms of construct validity by means of an EFA. The results yielded a unidimensional factor structure, with 19 items (all positively worded), and accounted for more than 60% of the variance in OL. This unidimensional factor structure was also confirmed with a CFA, and measurement invariance was also reported, by using the sector (public and private sector) as a grouping variable (Objective 2).

Based on these 19 items, a multidirectional OLS (with six negatively worded items) was developed and validated (Objective 3). The EFA yielded a two-factor model, also confirmed by means of a CFA, but because of estimation errors, a secondary factor model could not be assessed. It was subsequently decided to conduct a bifactor analysis, which produced the best fit statistics.

The bifactor model entails that all 19 items (although multidirectional) reflect a general factor, namely OL and two distinct sub-factors (constructive and destructive OL, respectively).

The two sub-factors of OL can be best described as:

Constructive organisational leadership – Leadership that integrates emotional intelligence, ethical integrity and strategic vision to inspire, empower and unite team members towards shared organisational goals. This dimension is characterised by self-awareness and emotional regulation, transparent and empathetic communication, high ethical standards and a focus on fostering individual and collective growth through meaningful relationships, clear direction and collaborative energy. Authentic transformational leaders motivate through their actions, challenge others to grow through engaging dialogue and create a supportive, trust-based environment that drives sustainable success.

Destructive organisational leadership – Detrimental leadership approach characterised by a lack of empathy, integrity and inclusivity, which undermines team morale and organisational success. This style is marked by dismissing diverse perspectives without compassion, acting without ethical principles, neglecting to coach or mentor team members for growth, failing to foster a safe emotional environment, rejecting diversity and neglecting to cultivate a sense of belonging or unity. Toxic leaders create divisive, unsupportive and disengaged workplaces, leading to diminished trust, poor collaboration and hindered individual and team performance.

Both versions, as a unidimensional (measured by unidirectional items) factor model and as the bifactor model (measured by multidirectional items), consisting of a general factor (OL) and two sub-factors (constructive and destructive OL), reported acceptable psychometric properties, including Cronbach alpha coefficients, CR, AVE, MSV and ASV.

The convergent validity of OL, as measured by both versions of the OLS, was further assessed by comparing it statistically to cognate relational leadership constructs. This includes authentic, ethical, servant and transformational leadership. Strong positive correlations were reported (regardless of the factor permutation), supporting the notion of convergent validity.

The fourth objective was to compare the response patterns of the two versions as well as the impact the directional differences have on their relationship with cognate leadership variables (using the same assessment as with Objective 3).

Note, a major difference at the item level in the response patterns was reported, although participants in Study 1 (unidirectional items) perceived their leaders more positively, with the most significant differences emerging in the reverse-coded items in Study 2. These items were mainly related to empathy, diversity, emotional safety and belonging. The assessment of the impact of the different versions on the relationship with other external (in this case cognate) leadership constructs yielded unconvincing results.

In conclusion, this study provides two equivalent versions of an instrument to measure OL, which is context-specific and was rigorously assessed. The first version is a unidimensional, positively worded item instrument measuring OL. The second is a multidirectional item instrument, consisting of two sub-factors, namely constructive and destructive OL, with a general OL factor (in accordance with the bifactor model).

The choice between the instruments could be influenced by various aspects, such as the nature of the population and sample (level of English proficiency, preparedness to participate in the study), the organisation and the rationale for the use of the instrument, the relationship between the participants and their leaders, etc. The choice of the multidirectional item might mitigate acquiescence bias, enhance content validity, reduce the risk of common method bias and improve measurement balance. It may however reduce consistency because of participant confusion or misinterpretation because of cognitive demand, and for the researcher, it requires careful handling of reversed items, complicating the analysis.

If the priority of the application of the instrument is clarity, reliability and ease of interpretation, and the population may be sensitive to complexity, it is recommended that the unidirectional item version be used. On the contrary, if the priority is detecting and correcting for bias, and the participants can handle cognitively demanding tasks, then the multidirectional item version is more appropriate.

Practical, methodological and managerial implications and the contribution of the study

This study contributes on practical, methodological and managerial levels. Practically, it provides two equivalent, reliable and valid versions of an instrument for measuring OL, within the SA context. Researchers can decide which version to use based on various aspects related to their studies, including the population, sample (for instance, educational level, English proficiency, etc.), the organisational realities (e.g. association of respondents with leaders), as well as the rationale of the study.

Methodologically, the study contributes by transforming a unidirectional item instrument into a multidirectional one and evaluating the effects of this change on the construct structure level as well as response patterns. Additionally, it incorporates bifactor analysis, an approach rarely used in leadership research, adding to the methodological contribution.

Managerially, the study enhances understanding of OL as either a unidimensional construct or one comprising two distinct sub-factors, namely constructive and destructive OL. This conceptualisation can inform leadership development programmes and coaching to better grasp OL’s nature in the South African context.

Limitations of the study and suggestions for future research

The primary limitations of this study stem from its methodology. Both versions of the OLS rely on self-reporting, which may introduce method bias, particularly in the unidirectional item version. Additionally, the cross-sectional design is a limitation, as it may artificially inflate the relationships between items and constructs, providing limited insight into how these relationships evolve over time or the direction of causality. The use of multisource validation of this instrument by means of multitrait-multimethod techniques (e.g. subordinate and peer rating) is also recommended.

This study is primarily descriptive and confirmatory, and it is recommended that the instrument be applied for additional conceptual and theoretical validation, reweaving the Afrocentric theoretical lens, and also in global settings, particularly in other collectivistic cultures.

Future research should further aim to explore OL’s impact on organisational outcomes such as climate, turnover intention and employee engagement.

Acknowledgements

This article includes content that partially overlaps with research from Dr Eben Enslin’s Doctoral thesis, titled ‘The conceptualisation, development, and validation of a South African organisational leadership scale’, submitted to the Graduate School of Business Leadership, University of South Africa, in 2023, under the supervision of Professor Anton Grobler. The thesis was part of a structured research programme led by Professor Grobler, who served as the primary researcher and data owner. For this journal publication, portions of the data, analysis and/or discussion from the thesis have been revised, updated and supplemented with an additional dataset. The original thesis is publicly accessible at https://go.exlibris.link/CNl89kRt. The authors confirm that this submission adheres to ethical standards for secondary publication, with appropriate acknowledgement of the original work.

The authors declare that they have no financial or personal relationships that may have inappropriately influenced them in writing this article.

Anton Grobler: Conceptualisation, Formal analysis, Methodology, Project administration, Supervision, Writing – original draft. Eben Enslin: Investigation, Methodology, Project administration, Writing – original draft, Writing – review & editing. All authors reviewed the article, contributed to the discussion of results, approved the final version for submission and publication, and take responsibility for the integrity of its findings.

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

The data that support the findings of this study are available from the corresponding author, Anton Grobler, upon reasonable request.

The views and opinions expressed in this article are those of the authors and are the product of professional research. It does not necessarily reflect the official policy or position of any affiliated institution, funder, agency or the publisher. The authors are responsible for this article’s results, findings and content.

References