Initial research efforts into intelligence, given the application of a tapered perspective, tended to associate intelligence solely with academic settings (Ang, Van Dyne & Tan, 2011). A growing interest in intelligence, however, has resulted in its classification per a range of foci such as emotional, social and general mental ability (Zhang, 2012). These intelligences account for most of the variations between the achievement levels of persons in the personal, social and work domains within their own cultures (Viggiano, 2016).

To fully explain intelligence, however, Sternberg and Grigorenko (2006) submit that cultural context must be considered. It is against this background that the concept of CQ was originally conceptualised and has evolved accordingly. CQ is founded upon the Sternberg and Detterman (1986) multiple loci of intelligence framework (Ang et al., 2011; Eisenberg et al., 2013; Kurpis, 2012; Ng et al., 2009; Peng, Van Dyne & Oh, 2015; Zhang, 2012). Although related to cognitive intellect as well as emotional and social intelligence (Ang et al., 2015), CQ situates discretely (Ang et al., 2007) largely due to the aforementioned intelligences being culture specific (Thomas et al., 2008). Contrasting them, CQ depicts an etic perception of intelligence as it transfers across cultures (Klenke, 2009; Ng & Earley, 2006).

Culturally intelligent persons are competent and effectual in multiple intercultural interactions as opposed to just a single or a few such exchanges (Ang & Inkpen, 2008; Ang et al., 2015). Thomas (2006) notes that CQ depicts the capability to not simply adjust to but actually influence the cross-cultural exchange. Accordingly, CQ helps to explain why some leaders easily adjust their perspectives and behaviours across cultures whilst others do not (Van Dyne, Ang & Livermore, 2010). Although various scholars, such as Fung and Lo (2017) as well as Thomas and Inkson (2005), have offered their own conceptualisation of CQ, that which was advanced by Earley and Ang (2003) has proven most popular in underpinning CQ research (Ott & Michailova, 2016). In terms of this last-mentioned model, CQ is theorised as possessing cognitive, motivational and behavioural components. A fourth component, metacognition, was subsequently added by Ang and Van Dyne (2008). The study on which this article reports made use of the Earley and Ang (2003) and Ang and Van Dyne (2008) CQ conceptualisation.

The four components or dimensions of CQ reflect different competencies (Ng, Van Dyne & Ang, 2012). Metacognition refers to higher-order cognitive procedures (Charoensukmongkol, 2016), that is, the processes through which persons source and grasp knowledge (Ang & Inkpen, 2008) and, hence, reflects the ability of a leader to strategise when traversing cultures (Van Dyne et al., 2010). As such, metacognitive CQ stimulates the formation of novel heuristics for intercultural interfaces and promotes the questioning of one’s own cultural suppositions (Ang et al., 2011). Cognition, the second of the CQ dimensions, represents a person’s knowledge about other cultures, examples of which include customs, standards and values (Van Dyne et al., 2012). This dimension of CQ, thus, signifies the extent to which a leader comprehends how to engage with others cross-culturally (Van Dyne et al., 2010). Ang et al. (2011), though, warn that unless the know-how emanating from cognitive CQ is assimilated into the other CQ dimensions, it is possible that such knowledge may not be of much value to leaders and could indeed be harmful. A leader, for example, with relevant cultural insights who lacks the desire (motivation) to display pertinent actions (behaviours) based on these might be viewed adversely by subordinates. The third dimension, motivational CQ, portrays the leader’s desire to acclimate interculturally (Van Dyne et al., 2010), that is, the energy expended in both acquiring knowledge about other cultures and immersing oneself in cross-cultural interfaces (Ng et al., 2012). Finally, behavioural CQ denotes the adoption and display of culture-appropriate actions (Ang & Van Dyne, 2008) that may be both verbal and non-verbal (Van Dyne, Ang & Nielsen, 2008). Ang et al. (2007) indicate that the dimensions ‘may or may not correlate with each other’ (p. 338), whilst Magnusson, Westjohn, Semenov, Randrianasolo and Zdravkovic (2013) argue that the dimensions are comparatively independent. In combination, the dimensions reflect the understanding that CQ concerns the integration of knowledge about cultural disparities with the impetus and competence to consider one’s own and other persons’ cultural programming, combined with the ability to display culturally apt behaviours (Schreier & Kainzbauer, 2016).

Although an abundance of leadership theories and styles exists (Avolio, Walumbwa & Weber, 2009; Brauckmann & Pashiardis, 2011; Esen, 2015; Landis et al., 2014; Perkins, 2013), leadership is most often demarcated according to leader traits, qualities and behaviours (Horner, 1997). Concentrating on leader behaviours, Pearce et al. (2003) defined a typology consisting of the transactional, transformational, empowering and directive leadership styles. Transactional leadership is considered by Golla and Johnson (2013) to be a style in which leaders promote employee self-interest through the promise of rewards in return for performance. Transformational leadership targets the attainment of organisational objectives by offering staff members a vision that eclipses their self-interest (Holten, Bøllingtoft & Wilms, 2015). Although the transactional and transformational leadership paradigm has attracted much scholarly attention (Clark & Waldron, 2016; Sims, Faraj & Yun, 2009), Hmieleski and Ensley (2007) insist that such leadership styles are often reflective of ambiguous behaviours. In contrast, empowering and directive leadership, on which empirical insights remain scanty (Kalaluhi, 2013; Sharma & Kirkman, 2015; Tekleab, Sims, Yun, Tesluk & Cox, 2008), are distinct from one another (Yun, Cox & Sims, 2006) and are situated at opposing ends of an action-based spectrum (Faraj & Sambamurthy, 2006). This research focussed on these last-mentioned leadership styles primarily because empirical investigation of them is limited. Further, as noted above, the transformational leadership style has dominated attention in the domain of leader CQ and there is some uncertainty as to the distinction between transformational and transactional leadership behaviours in the literature.

In the light of heightened global economic rivalry and with a view to improving their agility, many organisations have adapted their structural composition to reflect a more empowered workforce as opposed to the hierarchical leadership arrangements of yesteryear (Arnold, Arad, Rhoades & Drasgow, 2000). Empowering leadership aims to grow the capacity of followers to lead themselves (Mohamed, 2016; Tekleab et al., 2008) and may be defined as behaviours that promote power equality with staff members (Amundsen & Martinsen, 2014) or as the assignment of authority and responsibilities to subordinates (Hakimi, van Knippenberg & Giessner, 2010).

Empowering leadership embraces divergent employee viewpoints (Doh & Quigley, 2014) whilst spotlighting employee self-control advancement (Galanou, 2009). Ahearne, Mathieu and Rapp (2005) note that leader empowering behaviours consist of (1) enhancing work meaningfulness, that is, the extent to which a leader sets stimulating goals for subordinates and clarifies their contributions to organisational success (Kwak & Jackson, 2015), (2) encouraging decision-making involvement or the extent to which the leader facilitates decision-making by subordinates (Kwak & Jackson, 2015), (3) voicing sureness of high performance by, for example, attesting to subordinates’ capabilities (Kwak & Jackson, 2015) and (4) removing bureaucratic constraints or the degree to which a leader dissipates organisational factors that inhibit subordinates’ independence (Kwak & Jackson, 2015).

The empowering leadership style influences psychological empowerment positively (Amundsen & Martinsen, 2015; Joo, Park & Lim, 2016; Raub & Robert, 2012). It prompts followers’ perceptions of increased work challenges (Esteves & Lopes, 2017) and stimulates their feelings of self-efficacy (Biemann, Kearney & Marggraf, 2015; Bobbio, Bellan & Manganelli, 2012; Kim & Beehr, 2017). It demonstrates a positive relationship with employee creativity (Harris, Li, Boswell, Zhang & Xie, 2014; Zhang & Zhou, 2014) and innovative behaviour (Gkorezis, 2016). Empowering leadership also correlates with employees’ organisational job embeddedness (Erkutlu & Chafra, 2015) as well as with their work effort, performance and service delivery levels (Govender, 2016). It increases the extent of knowledge sharing between team members (Lee, Lee & Park, 2014) and change-orientated organisational citizenship behaviours (Li, Liu, Han & Zhang, 2016). The strengthening of relationships with managers and colleagues (Esteves & Lopes, 2017) as well as the advent of informal leadership (Adeel & Pengcheng, 2016) are also associated with this style of leadership.

Directive leadership, on the other hand, is based upon positional power (Lorinkova, Pearsall & Sims, 2013; Yun et al., 2006) and depicts behaviours that provide subordinates with precise direction concerning objectives, how such objectives must be realised and the outputs required (Martin, Liao & Campbell, 2013). As such, it embodies behaviours that are mostly aligned with task accomplishment (Dewettinck & van Ameijde, 2011). Accordingly, those leaders who adopt this style deliver leadership through the issuing of orders, commands and directives (Sims et al., 2009) based primarily upon their own judgements (Yun, Cox, Sims & Salam, 2007).

Directive leadership has been associated with some negative outcomes such as reduced team cohesiveness (Wendt, Euwema & van Emmerik, 2009) and slower product development (McDonough & Barczak, 1991). It does, however, influence employee role clarity positively (Dolatabadi & Safa, 2011). This style of leadership is also positively associated with a reduction in job demands that impede workers in the execution of their duties (Esteves & Lopes, 2017). Furthermore, it has a positive relationship with organisational commitment (Somech, 2005) and correlates with the quality of group results (Sagie, 1997) and processes (Peterson, 1997). Directive leadership is also an appropriate leadership style when subordinates lack requisite skills (Muczyk & Reimann, 1987), a major calamity is encountered (Maggitti, Slay & Clark, 2010) or the work activities are structured and routine (Sauer, 2011).

As pointed out in the introduction, transformational leadership appears to be the style that has attracted the most attention in the domain of CQ. Leader CQ, as an aggregate construct, cognitive CQ and behavioural CQ have all been confirmed as coinciding with transformational leadership (Ismail et al., 2012; Keung & Rockinson-Szapkiw, 2013). In addition, Ismail et al. (2012) found that transformational leadership coincided with both leader metacognitive and motivational CQ. Leader CQ also acted as a moderator of the transformational leadership relationship with both expatriate adjustment and performance (Lee et al., 2013) as well as organisational innovation (Elenkov & Manev, 2009). The relationship between the autocratic, democratic and laissez-faire leadership styles and the leader CQ dimensions was investigated by Eken, Özturgut and Craven (2014). Apart from reporting a positive relationship between leader motivational CQ and the democratic leadership style, these researchers were unable to detect any other significant associations. The authors are not aware of any empirical studies that have specifically concentrated on the relationship between leader CQ and both the empowering and directive leadership styles.

The following null hypotheses were thus set:

H1₀: There is no statistically significant relationship between the CQ (as a composite value) and the empowering leadership style of leaders at organisations operating in South Africa.

H2₀: There is no statistically significant relationship between the CQ (as a composite value) and the directive leadership style of leaders at organisations operating in South Africa.

H3₀: The relationship between leader CQ (as a composite value) and the empowering leadership style does not differ from that between leader CQ (as a composite value) and the directive leadership style.

H4₀: There is no statistically significant relationship between each of the CQ dimensions and the empowering leadership style of leaders at organisations operating in South Africa.

H5₀: There is no statistically significant relationship between each of the CQ dimensions and the directive leadership style of leaders at organisations operating in South Africa.

H6₀: All the leader CQ dimensions do not contribute uniquely and significantly in predicting the empowering leadership style of leaders at organisations operating in South Africa.

H7₀: All the leader CQ dimensions do not contribute uniquely and significantly in predicting the directive leadership style of leaders at organisations operating in South Africa.

The population consisted of all leaders (i.e. anyone to whom another staff member directly reports) at all organisations operating in South Africa. Conway (2000) and Kim and Yukl (1995) draw attention to the benefits of employing subordinate ratings as opposed to leaders self-reporting, whilst Ang et al. (2015) highlight the advantages of using informant-based CQ measures. Hence, sample data on the leaders were sourced from their subordinates.

The sample respondents, being the subordinates of the leaders, were recruited by students pursuing a Master of Business Leadership (MBL) degree through the Graduate School of Business Leadership at the University of South Africa (GSBL). The students, by means of their respective employers, gained access to the respondents. The students then acted as fellow researchers collecting data from 18 different South African organisations.

The corresponding author also collected data from the organisation by whom he is employed. In aggregate, data were collected from 1140 respondents across the 19 organisations. Six of the organisations trade within the manufacturing industry whilst another three operate within the telecommunications industry. Three of the organisations represent the media industry and a further two fall within the public sector. The defence industry was represented by a single organisation as was the banking sector. The remaining organisations were from the services industry. Although the respondents were selected on a random basis, from the personnel records in the participating organisations, the sample was based upon convenience as the choice of organisations was not random.

The following instruments, with the permission of their respective authors, were used to measure the variables and were presented to the respondents in the form of a single consolidated questionnaire. Details of the instruments appear below and are arranged per variable.

Leader CQ: Measured by the Cultural Intelligence Scale (CQS) devised by Van Dyne, Ang and Koh (2008) who note that both the self and observer report versions of the CQS are convergent, discriminant and criterion valid. Rockstuhl, Seiler, Ang, Van Dyne and Annen (2011) further point out that the CQS is reliable, based on Cronbach’s coefficient alphas (that ranged from 0.71 to 0.85) achieved by Van Dyne et al. (2008), and may be generalised across samples, time and cultures. This research made use of the observer report version.

Empowering leadership: Evaluated via the 10-item instrument of Ahearne et al. (2005). Yoon (2012) confirmed the reliability of this instrument (Cronbach’s coefficient alpha of 0.93), whilst Zhang and Bartol (2010) remark that it depicts the distinctiveness of empowering leadership.

Directive leadership: Assessed through six items developed by Pearce and Sims (2002) and four from Hwang et al. (2015). Hinrichs (2011) states that the items developed by Pearce and Sims (2002) have delivered acceptable levels of reliability (Cronbach’s coefficient alpha of 0.88), whilst Hwang et al. (2015) note a similar result in respect of the items they developed (Cronbach’s coefficient alpha of 0.85).

Following receipt of permission from the GSBL Research Ethics Review Committee for the data to be collected, participating MBL students were fully briefed on the nature, purpose and importance of the research. The questionnaire as well as the respondent information sheet were explained to them. Each student first sourced the written approval of his or her employer’s chief executive officer, or other appropriate executive, on the basis that it would not be named. Thereafter, a list of potential respondents (i.e. subordinates of the leaders) was obtained from the human resource department. Each name on the list was allocated a number. Potential respondents were then selected randomly using a random number generator. Chosen persons were invited to a meeting at which the research was introduced to them. They were advised that participation was completely voluntary and anonymous. Those staff members who agreed to participate were handed a hard copy of the questionnaire and were requested to complete it at the meeting, after which they returned it to the respective student. The reason why a hard copy, rather than a soft copy, of the questionnaire was used is because it facilitated completion during the aforementioned meeting rather than having to follow up with the respondents (potentially on numerous occasions) following the meeting. An Excel-based template was then provided by the authors to the students wherein each captured the details from the questionnaires they had collected. The populated templates were then consolidated. The corresponding author followed the same research procedure in collecting the data from his employer.

IBM SPSS (version 24) (Field, 2012) was used to perform the data analysis. Frequencies were calculated to provide a descriptive view of respondent demographics. Measures of central tendency were also computed for the different variables included. Cronbach’s coefficient alphas were calculated to assess reliability. Reliability was accepted as satisfactory where the alpha scores exceeded 0.70, with scores above 0.80 being taken as desirable (Pallent, 2011).

Validity was examined through principal axis factor analysis with direct Oblimin rotation. The number of factors retained was based upon the rule of thumb that only those factors with eigenvalues larger than one would be included (Coovert & McNelis, 1988). Factor numbers were also confirmed through an examination of scree plots, as recommended by Costello and Osborne (2005). The process, as described by Courtney (2013), for such examination was followed; that is, the scree plot was scanned for a break or hinge (also referred to as an ‘elbow’). The number of factors was considered to be the number of eigenvalues appearing before the elbow. Factor acceptability was evaluated with reference to the guidelines of Hair, Black, Babin and Anderson (2010) and Osborne and Costello (2009), which indicate that item loadings of 0.30–0.40 are minimally acceptable whilst loadings of 0.50 and greater are favoured. Although factors that have at least four adequately loading items are generally noted to be acceptable (Osborne & Costello, 2009), there is little guidance in the literature as to the suitable number of items that should load when measurement scale length is considered. Hence, for a factor to be accepted in this research, it was decided that at least 80% of the respective items measuring it should load with a minimum weight of 0.50 each.

Pearson product-moment correlation coefficients were calculated to gauge the relationships between the variables. Only statistically significant correlations were considered. Significance was taken at the 5% level as recommended by Lazaraton (1991). To assess the practical significance of the alphas, the guidelines of Cohen (1988) were followed; that is, correlation coefficients above 0.10 were accepted as small, those exceeding 0.30 were appraised as medium and those greater than 0.50 were taken as large. The online calculator of Lee and Preacher (2013), which is based upon the Fisher r-to-z transformation and the asymptotic equations of Steiger (1980), was used to evaluate the difference between correlations.

The ability of the leader CQ dimensions to predict empowering and directive leadership was determined through stepwise regression analysis. Only those dimensions that contributed statistically significant predictions were considered. The f² statistic was used in judging the practical significance of the models; in this regard, the conclusions of Ellis and Steyn (2003) based on the guidelines of Cohen (1988) were followed. According to Ellis and Steyn (2003), where f² < 0.15, the effect is small and R² is not practically significant, where 0.15 < f² < 0.35, the effect is medium and R² is practically significant and where f² > 0.35, the effect is large and thus R² is practically important. Emphasis was placed on those betas that contributed uniquely and independently to the variance in the dependent variable.