Engaged employees apply themselves totally (physical, mental and emotional) to their job role (Kahn, 1990). Kahn identified three states that employees must pass through before achieving full engagement. These are psychological meaningfulness, availability and safety. Psychological meaningfulness occurs when individuals perceive that they are obtaining equal or greater benefits compared to their investment of physical, mental and emotional efforts in the work role. Employees are psychologically safe when they perceive that they can express themselves in their work without negative consequences towards self-image, career or status (Kahn, 1990). Psychological availability refers to the perception that one has enough physical, emotional, and mental resources to engage in work at any moment in time (Kahn, 1990). Engaged employees make a valuable contribution to the productivity and image of the organisation and have positive work attitudes (Anikan & Oyewole, 2014; Harter, Schmidt, & Hayes, 2002; Ugwu, 2013; Young, 2014). The three main drivers of employee engagement are leadership style, organisational climate and individual dispositions (Anikan & Oyewole, 2014; Ugwu, 2013). Commenting on leadership as the main driver of engagement, Young (2014) stated that leadership behaviour holds the key to the improvement of engagement and that the correlation between leadership and engagement is as high as 0.6. Kruse (2012) had earlier itemised that the leadership behaviours that affect engagement include communication, growth and development of employees, reward and recognition, trust and confidence, and emotional intelligence.

Employee engagement is an outcome of good employment relations, driven by the behaviour of leaders in the organisation (Koene, Vogelaar, & Soeters, 2002). Leadership styles have foundational motives and such motives are reflected in the leadership behaviours. For example, the motivation behind the AL and TL styles is fuelled by the inward mindset and interests of the leader rather than those of the employees. Both leadership styles see employees as means to an end (Lavinson, 1980). The success of TL is hinged on the leader’s ability to appeal to the follower’s self-interest (Lavinson, 1980). Hence, it is argued that TL behaviour cannot drive employees to put in a discretionary effort to help the organisation succeed, as employees concentrate on meeting only the agreed objectives (Levinson, 1980). Autocratic leaders maintain absolute authority over employees, demand unquestionable obedience, practise unilateral decision-making and create a widening power distance between them and employees (Cheng, Chou, Wu, Huang, & Farh, 2004). In doing these, the leader achieves full control over employees to achieve desired goals (Schuh, Zhang, & Tian, 2013). The widening power distance is viewed by employees as apathy towards their own interests. For example, an autocratic leader will ignore employees’ suggestions and play down their contributions towards the success of the organisation (Aryee, Chen, Sun, & Debrah, 2007). Thus, the environments created by transactional and autocratic leaders are characterised by one-way communication, based purely on exchange principles with no emotional attachment, devoid of trust and confidence, low emotional intelligence and poor development of employees (Hollander, 1986). Such an environment is de-energising to employees. By the theory of social exchange and norm of reciprocity, employees in such environments will not invest in their work role by being engaged as the cost is more than what they get out of work. This violates the principle of psychological meaningfulness. Hence, the following hypotheses are postulated:

The motive for SL is to serve the employees so that they give their best to achieve organisational productivity. The servant leader’s priority is to build relationship with his followers through joint decision-making as a means of getting employees to engage in activities that will benefit the organisation (Gregory Stone et al., 2004; Savage-Austin & Honeycutt, 2011). Thus, the success of SL lies in employees’ willingness to provide a discretionary effort as opposed to achieving only what is expected of them. Such leadership behaviour will encourage two-way communication, emotional attachment between leader and follower, build trust and confidence, and help employees to develop high emotional intelligence. Employees benefiting from these positive behaviours will reciprocate by way of high employee engagement. The three principles, namely, psychological meaningfulness, safety and availability are satisfied in the environment created by servant leaders. Based on the arguments above, the following hypothesis is postulated:

Employee voice is described as the discretionary verbal communication of ideas, suggestions or opinions by employees with the intention of improving individual and organisational performance (Morrison et al., 2011). Leadership attitude and beliefs on EV will affect the tendency for employees to make suggestions on improving organisational outcomes (Amah & Okafor, 2008a; Gul, Ahmad, Rehman, Shabir, & Razzaq, 2012). In leadership behaviours that do not encourage voice, employees will perceive high cost to themselves and career if they voice out issues about the organisation, especially those perceived as negative by such leaders (Amah & Okafor, 2008a; Vakola & Dimitris, 2005). Servant leadership style values employees and develops them to their full capability. It allows EV in whatever is important in the organisation as its essence is the full development of employees (Amah & Sese, 2018). Autocratic style is based on one-way communication in which the leader tells the employees what to do, how to do it and when to do it, and autocratic leaders believe that employees are meant to follow orders and not to offer suggestions (Cheng et al., 2004). Hence, the climate created by this style will not favour EV. Transactional leadership style values employees only on exchange basis. It will allow EV provided it is meant to pursue the interest of the leader. Employee voice may be encouraged, but employees may also not feel very comfortable using such voice for fear of consequences (Amah & Sese, 2018). Engaged employees are always seeking to improve efficiency in their job, hence the more voice they are allowed, the higher they are engaged (Amah & Sese, 2018). Highly engaged employees are always identifying ways to improve personal and organisational productivity (Amah & Sese, 2018). In an environment in which EV is encouraged, engaged employees will utilise voice to make suggestions aimed at improving the work process and to enhance their productivity (Budd & Zagelmeyer, 2010). By the theory of social exchange and norms of reciprocity, such employees have high level of engagement. Hence, the following hypotheses are postulated:

Amah and Okafor (2008b, p. 494) define POS as, ‘the perception of employees that the organisation values their contribution’. An environment of high POS is created by leadership behaviours that value and care for people. A social exchange theory has always been used to explain how POS affects employees’ work outcome, including work attitudes (Eisenberger, Huttington, Hitchinson, & Sowa, 1986). The servant leadership style creates an environment that values people and provides them with the means to be effective. The style will create a climate which will be perceived by employees as supportive. On the contrary, both AL and TL styles do not value employees and thus will create environments that will be perceived as unsupportive (Amah, 2017). Highly engaged employees will perceive high POS as a confirmation that the organisation values the personal sacrifices they make in being highly engaged. Hence, such employees will feel obligated to make a more discretionary effort to benefit the organisation. Greater perceived organisational support is expected to produce greater affective emotional attachment and feelings of obligation towards the organisation based on social exchange theory and norms of reciprocity (Allen, Shore, & Griffeth, 2003). Hence, the following hypotheses are postulated:

The model tested is shown in Figure 1.

The study was based on a cross-sectional survey design and quantitative method of analysis. Data were acquired through questionnaires administered to randomly selected participants in organisations located in the city of Lagos, Nigeria. The questionnaires were administered between the period of July and October 2017. The organisations came from three industries, namely, oil, financial and manufacturing.

The human resource managers from each organisation reviewed the questionnaires prior to distribution to ensure that no confidential information was involved and that no ethical issue was violated. A research assistant worked with designated individuals from each organisation to distribute the questionnaires directly to the randomly selected participants in each organisation. The filled-out questionnaires were returned in sealed envelopes provided by the research assistant. Participants were assured that confidentiality would be maintained and that the ultimate results would not be given to their organisations in a format that would reveal their identities.

Analysis of data were carried out using SPSS 23 and Analysis of Moments Structures (AMOS 25). Model fit was judged for goodness of fit index (GFI), comparative fit index (CFI) and root-mean-standard-error of approximation (RMSEA). For a model to be described as having good fit, CFI and GFI must be 0.9 and above, and RMSEA should be between 0.05 and 0.08 (Hu & Bentler, 1999).

Six factors were extracted in an unrotated factor analysis, with the first factor extracting only 20% of the variance when compared with 73% extracted by the other five factors. Hence, common method variance was not an issue in the analysis (Podsakoff & Organ, 1986).

A condition for using structural equation modelling is that the distribution of the data must satisfy the multivariant normality test. The multivariant kurtosis obtained is 1.99 with critical ratio of 1.80 which is within the range of kurtosis and critical ratio of ±2.58 and ±1.96, respectively, recommended by Hair, Black, Babin, Anderson and Tatham (2010). Construct validity shows the extent to which the items for a factor load exclusively on the factor. The item loadings in Table 2 are all above the cut-off point of 0.5 which indicates acceptable construct validity (Hair et al., 2010). Convergent validity measures the level of correlation between the items that measure the same factor. Convergent validity is established when factor loadings exceed 0.5 (Hair et al., 2010). Cronbach’s alpha is more than 0.70 (Nunnally, 1978), and average variance extracted more than 0.5 (Hair et al., 2010). Table 2 indicates that these conditions are satisfied. Discriminant validity is established by comparing the variance extracted by each factor with the correlations the factor has with other factors (Koufteros, Vonderembse & Doll, 2002). Each diagonal element in Table 3 is greater than its vertical and horizontal off diagonal elements showing evidence of discriminant validity.

Descriptive statistics, Cronbach’s alpha coefficients and correlations for the study variables SL, AL, TL, EV, employee engagement and POS are presented in Table 4. The figures in the table indicate that the Cronbach’s alpha coefficients for all the study variables are above the acceptable value of 0.7. Construct, convergent and discriminant validities are adequate. Hence, the results obtained with the estimation method are reliable. All the study variables are positively correlated with employee engagement. The correlations involving the AL and TL styles are opposite of what was hypothesised. The means for the SL, AL and TL styles are 3.99, 4.25 and 4.47, respectively. This shows that fewer leaders exhibited SL behaviour in the sampled organisations.

After the initial estimation of the hypothesised model, modification index recommended a relationship between EV and POS. This relationship is implied in the work of Vakola and Dimitras (2005). Estimation of the new alternative model showed that the new model has better fit and this was confirmed by the χ² difference test. The estimated model is in Figure 2, while the goodness of fit statistics are in Table 5. None of the leadership styles has a significant direct relationship with employee engagement as hypothesised. Thus, Hypotheses 1a (β = 0.02, t = 0.751, p = 0.317), 1b (β = 0.09, t = 0.762, p = 0.446) and 1c (β = 0.06, t = 0.748, p = 0.443) are not supported. The relationship between AL and EV is not significant (β = 0.17, t = 1.911, p = 0.056), thus H2a is not supported. The relationship between TL and EV is positive and significant (β = 0.30, t = 2.277, p = 0.023), thus H2b is supported but in opposite direction. The relationship between SL and EV is significant (β = 0.36, t = 3.726, p = 0.001), thus H2c is supported. Employee voice is positively related to employee engagement (β = 0.21, t = 3.179, p = 0.01), thus H2d is supported. Servant leadership is positively related to POS (β = 0.56, t = 7.235, p = 0.001), thus H3a is supported. The relationship between AL and POS is not significant (β = −0.07, t = −1.062, p = 0.288), thus H3b is not supported. Transactional leadership is positively related to POS (β = 0.27, t = 2.797, p = 0.005), thus H3c is supported but in the opposite direction. Perception of organisational support is positively related to employee engagement (β = 0.33, t = 3.838, p = 0.001), thus H3d is supported. Employee voice is positively related to POS (β = 0.21, t = 3.179, p = 0.01). This relationship was not hypothesised but recommended by the modification index. The model explained 51% of the variance in employee engagement. Autocratic leadership, TL and SL have standardised indirect effects of 0.063, 0.24 and 0.37 on employee engagement, respectively (there were no significant direct effects). Thus, SL explained more variance in the value of employee engagement when compared with other leadership styles.