Psychological research has revolved around the behaviours of WEIRD (Western, Educated, Industrialised, Rich, and Democratic) people (Espinosa & Verney, 2020), and according to Grobler and Singh (2018) can also be applied to organisational (and leadership) research in general. Syed (2021) and Syed and Kathawalla (2020) highlighted the relative absence of race, ethnicity and culture in research, specifically identifying the exclusion of structural considerations (e.g. power, privilege and context) in understanding personality. In doing so, Syed (2021) emphasises that personality encompasses more than just traits and should include structural contexts and social structures. There have been calls for research that considers culture, race and ethnicity too in personality psychology, and this has been viewed as a much-needed avenue for scholars to engage with real-world issues (Arshad & Chung, 2022; Dupree & Boykin, 2021; Roberts et al., 2020). Cheung et al. (2011) emphasise the necessity to analyse a questionnaire thoroughly for comparability across different culture groups, before any decision or prediction can be made about the construct that was measured. Psychological science tended towards excluding people from underrepresented groups (Espinosa & Verney, 2020; Syed et al., 2018). The main focus of this research is on instrument transportability and the investigation of the psychometric properties of foreign-developed instruments when used on a South African sample.

Subsequently, the focus of this research was on the measurement of transformational leadership in a diverse South African context and the soundness of the Transformational Leadership Questionnaire (TLQ), developed by Podsakoff et al. (1990), from an etic perspective. This research explored the applicability of a predominantly Western construct and instrument in a context that is extremely diverse, specifically in terms of the cultural differences found in South Africa (Nel et al., 2012; Taylor, 2004).

The original TLQ, that consisted of 22 items, identified six factors, namely articulating a vision, providing an appropriate model, fostering the acceptance of group goals, high-performance expectations, individualised support and intellectual stimulation.

The universal utilisation and application of a non-South African-based instrument were investigated within the diverse South African context (etic approach). The research aimed to apply an existing instrument (TLQ) in a new context for a diverse sample to confirm the construct validity (Morris et al., 1999).

Our study has two broad objectives. Firstly, it aims to examine the TLQ in terms of construct (factorial) validity; an exploratory factor analysis (EFA) as well as confirmatory factor analysis (CFA) were used to conduct the analysis. Secondly, discriminant and convergent validity of the TLQ were investigated to assess the factorial validity of the instrument.

The contribution of this research establishes a valid measure of transformational leadership from an etic approach. The necessity of this research originates from the diverse nature of the South African workforce. Participants from both the private and public sectors participated in our study.

An EFA was conducted to determine the construct validity of the TLQ (from an etic approach). The EFA was conducted on the 22 items of the TLQ using the data from two studies independently, which had 2034 and 1771 respondents, respectively. Prior to running the analysis with SPSS (version 25), the data were screened by examining descriptive statistics on each item and possible univariate and multivariate assumption violations. From the initial assessment, all variables were reported to be interval-like, variable pairs appeared to be bivariate, the responses were normally distributed, and all cases were independent of one another.

The relatively large sample size contributed to adequate variable-to-case ratios (92:1 and 81:1, respectively, with the combined data set being 173:1).

In order to assess the appropriateness to conduct an EFA, the KMO measure of sampling adequacy and Bartlett’s test of sphericity were performed. The results are reported in Table 4.

The value of the KMO measure of sampling adequacy exceeded the critical cut-off value of 0.60 supporting the strategy to perform an EFA. The results of the Bartlett’s test of sphericity yielded Chi-square values of 28 599.89, 22 739.35 and 50 128.48 (degrees of freedom [df] = 231; all statistically significant at p ≤ 0.001). These results are an indication that an EFA is an appropriate statistical technique to assess and evaluate construct validity. The K1 rule was further used with the scree plot to determine the number of factors to retain. The results of the K1 rule for each of the studies independently as well as the combined data set are reported in Table 5.

The factor solutions (with eigenvalues close to or larger than one) reported in Table 5 consist of two factors in Study 1 and three factors in Study 2. The total variance explained in the two studies as well as the combined data set is close to 60%. Cattell’s scree test, which is aimed at retaining the components (factors) before the break (elbow rule), was performed on all three data sets independently and the results are reported in Figure 1.

The interpretation of the scree plot, as depicted in Figure 1 was problematic, as there was one strong factor, but it is not clear in any of the three figures where the elbow flattens off.

The process to identify the correct number of factors is, according to Pallant (2020), further complicated by the conservative nature of the K1 rule as well as the scree plot.

Subsequently, the Monte Carlo parallel analysis simulation technique, which is regarded to be a more stringent technique was employed. The Monte Carlo parallel analysis aims to determine the number of factors that account for more variance than the components derived from random data. The eigenvalues obtained from the actual data are compared with the eigenvalues obtained from the random data. If the actual eigenvalues from the principal component analysis of the actual data are greater than the eigenvalues from the random data, the component (factor) is retained. Four components were included in the Monte Carlo parallel analysis to test the assumption of a two-component (factor) solution, as suggested by the interpretation of the K1 rule (Table 5). The results are reported in Table 6.

The results of the Monte Carlo parallel analysis yielded a two-component (factor) model, confirmed for the two independent studies as well as the combined data set. The results of the correlational analysis (Pearson correlation) between the two retained components, from now on referred to as the TLQ factors, are reported in Table 7.

Relatively high, positive correlations (ranging between 0.56 and 0.80) between the TLQ factors were reported for the two independent studies, as well as the combined data set. All the correlations thus constitute a medium to large effect and practical significance ranging between 0.56 and 0.80, supporting the notion of convergent validity (Cohen et al., 2013). The structure coefficients from the Promax rotation are reported in Table 8.

The results of the factor analysis conducted on the TLQ are summarised in Table 8. As proposed by Meyers et al. (2013), the inclusion criteria for an item was a factor loading cut-off point of 0.40. The factor solution differs from the original 5 factor solution with 22 items of Podsakoff et al. (1990). Eighteen of the 22 items of the TLQ loaded on the two factors. The items excluded (items 7, 12, 14, 15) were analysed but no clear reason (based on theoretical alignment and language, vocabulary considerations) could be found for their exclusion (except a relatively low factor loading). It was decided to continue with the 18 items that met the criteria.

The items loaded in a quite different manner compared with the original configuration reported by Podsakoff et al. (1990). The communalities of the items, across the two studies, although not reported in Table 8, are relatively high, ranging between 0.73 and 0.43, 0.75 and 0.46 and 0.74 and 0.50 for the two studies, respectively, as well as the combined data set.

Construct validity (specifically factorial validity) was assessed through an CFA. The results of the three models tested are reported in Table 9.

The three models were evaluated based on an inspection of the respective fit of the CFA. The worst fitting model was the one-factor model (all 18 items) (CFI = 0.98, RMSEA = 0.045) (model_a). The second-order factor model (model_b) is marginally better than the first-order factor model by analysing the chi-square test values. The difference in chi-square between the second-order factor and first-order factor models is 120 (i.e. 837 – 717), which is distributed as chi-square with 80 – 54 = 26 degrees of freedom. The second-order model (model_c) is thus the best fitting model. The 18 items loaded directly on their respective factors (i.e. Leadership focused on organisational, group and individual transformation and Leadership focused on creativity and innovation) which then contributes to a higher order or secondary factor, namely transformational leadership.

Furthermore, elementary cross-validation analysis was performed to assess the possibility of invariance between the two sample groups, namely the private and public sectors. The sample was split into the two sectors, using 500 cases randomly selected from each sector (from the total sample consisting of 2034 and 1771 respondents from the two sectors, respectively). The results reported for the two sample groups were χ²/df (237) = 2.97, CFI = 0.98, TLI = 0.97, RMSEA = 0.04, ECVI = 0.42 and χ²/df (243) = 3.05, CFI = 0.99, TLI = 0.97, RMSEA = 0.05, ECVI = 0.43 for the private and public sectors, respectively. The degree of invariance in terms of the Likelihood Ratio Test is 0.08 (3.05–2.97), and similar values were reported for TLI (0.97), which is lower than the norm of 0.05. The ECVI values for the private and public sectors are 0.42 and 0.43, respectively (difference = 0.01), which is marginal.

In addition to the comparison of the sectors, it was deemed necessary to determine invariance, using the management and non-management groups. The sample was divided into 1000 cases each for the management and non-management groups (randomly selected from the total sample). The results reported for the two sample groups were χ²/df (251) = 3.14, CFI = 0.99, TLI = 0.98, RMSEA = 0.04, ECVI = 0.34 and χ²/df (210) = 2.64, CFI = 0.99, TLI = 0.98, RMSEA = 0.04, ECVI = 0.39 for the management and non-management groups, respectively. The degree of invariance in terms of the likelihood ratio test is 0.52 (3.14–2.62) and similar values were reported for TLI (0.98). The ECVI values for the two groups are marginal with a difference of 0.05 (0.39–0.34). The comparison of the four sample groups by means of cross-validation, lend support to the notion of invariance, regardless of the level or role (management and non-management) of the participants.

In addition to the CR, AVE, MSV and ASV that looked at the convergent and discriminant validity in terms of the TLQ items that loaded on each of the factors, the convergent and discriminant validity of the two factors were also determined. The convergent validity of the TLQ factors was investigated by means of a heterotrait-mono method coefficient (HTMM coefficients) by comparing the TLQ factors and the total transformational leadership score to a range of cognate constructs.

All the selected instruments that measure the cognate constructs were evaluated (based on previous research) on their psychometric properties, specifically the Cronbach’s alpha coefficients (all reported α > 0.70), are: Authentic leadership (Walumbwa et al., 2008) (Studies 1 and 2); Employee engagement (Schaufeli et al., 2006) (Study 1); LMX or leader–member–exchange (Linden & Maslyn, 1998) (Studies 1 and 2); Organisational energy (Cole et al., 2012) (Study 2); Perceived organisational support (Eisenberger et al., 1986) (Study 1); Person-organisational fit, including Supplementary fit and Complementary fit (Cable & DeRue, 2002; Grobler, 2016a) (Studies 1 and 2); Psychological capital (Grobler & Joubert, 2018; Luthans et al., 2007) (Study 2); Psychological contract, on both the self and the organisation level (Freese & Schalk, 1997) (Studies 1 and 2); Pro-active work behaviour (Parker & Collins, 2010) (Study 2) and lastly Turnover intention (Brashear et al., 2003; Grobler & Grobler, 2016) (Study 1 and 2).

The Pearson product moment correlations are reported in Table 10.

It is apparent that there is some degree of convergence through the inspection of the correlations (all p ≤ 0.001) in Table 10. This is specifically true in terms of the hypothesised relationship between transformational leadership and its two factors, included in the table as F¹: Leadership focused on organisational, group and individual transformation and F²: Leadership focused on creativity and innovation, and related leadership constructs. Authentic leadership reported high correlations (r ranging from 0.54 to 0.71 for the factors and TFL total score), which is similar to that of LMX with r ranging from 0.52 to 0.74.

It was further hypothesised that transformational leadership would have a positive impact on organisational behaviour, which is true for Employee engagement in Study 1 (r = 0.38, 0.37 and 0.39) and Proactive work behaviour in Study 2 with r ranging from 0.27 to 0.39.

High correlations were found between the two TLQ factors and the interactionist concepts, with Person organisational fit ranging from r = 0.38 to 0.48. Psychological contract reported high correlations across the two studies with r ranging from 0.32 to 0.42. A relatively high, negative correlation between the TLQ factors and Turnover intention was reported (ranging from r = –0.22 to –0.37).

A further determination of discriminant validity was conducted through multiple regression, with the TLQ factors, Leadership focused on organisational, group and individual transformation and Leadership focused on creativity and innovation being regarded independent or predictor variables. The rationale for this analysis is to determine the uniqueness of the contribution (after inspection of the difference in the beta values), which is an indication of discriminant validity. The results are reported in Table 11 (standard errors are not included because of the limited space in the table and because of the fact that the matter of interest for discriminant validity is the difference in beta values).

All the multiple regression results, as reported in Table 11, are statistically significant (p ≤ 0.05). The results support the notion of discriminant validity, as the two TLQ factors (F¹: Leadership focused on organisational, group and individual transformation and F²: Leadership focused on creativity and innovation) contribute uniquely to the multiple regression models in terms of the variance accounted for in the dependent variables if the magnitude of beta values are compared.

The purpose of our study was to examine the instrument properties of the TLQ of Podsakoff et al. (1990), which consists of 22 items, and six factors (articulating a vision, providing an appropriate model, fostering the acceptance of group goals, high performance expectations, individualised support and intellectual stimulation), within the South African context. Our study comprises a well-representative sample across two independent studies. Our study adopts an etic approach, as the TLQ is an imported instrument that is now being scrutinised and adapted for use within a specific context, in this case South Africa. It is important that any measurement, specifically in such a diverse country such as South Africa, is validated for that specific context. Subsequently, our study is based on two broad objectives. The first objective is the determination of the construct (factorial) validity of transformational leadership, as measured by the TLQ. The second objective is to assess the discriminant and convergent validity of the TLQ.

Exploratory factor analysis and CFA were used to determine the construct validity and specifically the underlying structure using the original version of the TLQ (22 items). The use of an EFA was supported by the results of Bartlett’s test of sphericity and the KMO. A two-factor solution emerged and was retained after conducting the EFA with Promax rotation, explaining close to 60% of the variance (on average across the two studies independently, as well as in the combined study). The Monte Carlo parallel analysis simulation also supported the two-factor solution.

Acceptable psychometric properties were reported for the two factors (with α ranging between 0.79 and 0.95). A new naming convention that is descriptive of the two new factors was developed. The factors were named in accordance with the unique item composition, which differs significantly from the original instrument. The first factor, Leadership focused on organisational, group and individual transformation contributed to the most variance explained in transformational leadership with 48% and the 15 items that loaded on this factor are from all the original factors, and not in a specific order or pattern. The second factor, called Leadership focused on creativity and innovation contributed 8% to the explanation of the variance in the overall construct and consists of three items. Items from the original instrument that did not load on the factor structure were items 7, 12, 14 and 15. No substantive reason could be found for their exclusion.

The results of the EFA were used to conduct the CFA. The second-order model was identified as the best-fitting model. This second-order factor model consists of Transformational leadership as a super factor and equal contributions of the two factors (leadership focused on organisational, group and individual transformation and leadership focused on creativity and innovation) (χ²/df (80) = 5.30, CFI = 0.99, RMSEA = 0.034).

A rigorous investigation into the validity of the construct (and the two factors) was carried out, which included convergent validity. Convergent validity was assessed based on the hypothesised relationship it has with other, cognate constructs. The relational leadership constructs chosen were Authentic leadership and LMX. The organisational behaviour related constructs included are Organisational energy; Perceived organisational support; Person-organisational fit; Psychological capital; Psychological contract; Pro-active work behaviour, and lastly Turnover intention. The relatively high correlations between F¹: Leadership focused on organisational, group and individual transformation and F²: Leadership focused on creativity and innovation as well as the total transformational leadership score and the related measures is an indication of acceptable convergent validity.

Lastly, an assessment of discriminant validity was conducted through the use of multiple regression. The newly configured TLQ factors, F¹: Leadership focused on organisational, group and individual transformation and F²: Leadership focused on creativity and innovation were used as an independent or predictor variables, with the work attitudinal and positive organisational behaviour constructs as dependent variables. The purpose of this procedure was to determine whether there is a difference between the two factors (discriminant validity). The results of the multiple regression support the notion of discriminant validity, as the two factors reported unique contributions in the regression model (different beta values).

The value of our study rests on the adapted (reconfigured) TLQ instrument, validated for the use in the South African context. The adapted TLQ was found to be valid on structural (factorial) as well as external (discriminant or convergent) levels, and it was found to be reliable. The results suggest that this valid, adjusted measurement of transformational leadership could be used with confidence within the South African context. It further also emphasises the need to scrutinise instruments that were developed in other parts of the World before it is used in a different context. This is even more applicable on the diverse South African population.