A non-experimental cross-sectional correlational survey-based research design was used (Salkind, 2012) to determine the interrelationship between perceptions of leader empowering behaviour, psychological empowerment, organisational citizenship behaviours and turnover intention within a manufacturing division of an organisation. Salkind (2012) indicate that this research design is appropriate to sample behaviours, cognitions and affect at a current moment in time. This type of design is ideally suited to address the descriptive and predictive functions associated with a correlational design, in which relationships between variables are investigated (Shaughnessy, Zechmeister & Zechmeister, 2003). However, no causal inferences can be established or inferred (Salkind, 2012).

An availability sample (Salkind, 2012) of employees working in a manufacturing division was drawn. A total population of 300 employees working within the manufacturing division of the organisation was targeted, and 200 usable questionnaires were obtained. Descriptive information of the sample is provided in Table 1.

The study population consisted of mainly Afrikaans-speaking (38.0%) male employees (67.0%) between the ages of 25 and 35 years (37.5%). The majority of the sample had a Grade 12 level of education (43.5%), had two to five years of service within the manufacturing division (30.5%), and was permanently employed (81.4%). The sample was representative of the actual population of the manufacturing division within the company.

The Leader Empowering Behaviour Questionnaire (LEBQ) (Konczak et al., 2000) was used to measure employees’ perceptions of leader empowering behaviour. The instrument consists of 19 items, scored on a seven-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree). The scale measures six dimensions of leader empowering behaviour: delegation of authority (3 items), accountability for outcomes (3 items), self-directed decision-making (3 items), information sharing (4 items), skills development (3 items) and coaching for innovative performance (3 items). ‘My manager gives me the authority I need to make decisions that improve work processes and procedures’ is an example of self-directed decision-making. The original instrument consisted of 17 items, but two items on information sharing were added from Arnold et al. (2000). These items are: ‘My manager explains his or her decisions and actions to my work group’ and ‘My manager shares company goals to my work group’. Konczak et al. (2000) argues that a six-factor structure provides leadership with precise and detailed feedback and explains a greater percentage of the total variance. Konczak et al. (2000) report reliability coefficients that range between 0.82 and 0.88. Higher scores indicate higher perceptions of leader empowering behaviours.

TheMeasuring Empowerment Questionnaire (MEQ) (Spreitzer, 1995) was utilised to measure the level of psychological empowerment experienced by employees. The items are scored on a seven-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree). The scale consists of four subdimensions of psychological empowerment with three items each. Examples of items are ‘The work I do is very important to me’ (meaning); ‘I am confident about my ability to do my job’ (competence); ‘I can decide on my own how to go about doing my work’ (self-determination) and ‘My impact on what happens in my department is large’ (impact). Spreitzer (1995) reports the following alpha coefficients for the subscales: 0.92 for meaning, 0.90 for competence, 0.85 for self-determination and 0.84 for impact, with an overall Cronbach alpha coefficient of 0.92. According to Spreitzer (1995), a four-factor structure explains a large percentage of the total variance of the instrument, which indicates the construct validity for the instrument.

TheOrganisational Citizenship Behaviour Questionnaire (OCBQ) (Van Dyne, Graham & Dienesch, 1994) was used to measure the strength of employees’ organisational citizenship behaviours in five dimensions, namely loyalty, deviant behaviour, social participation, advocacy participation and functional participation. A seven-point Likert scale, ranging from 1 (strongly disagree) to 7 (strongly agree), is used. High scores indicate strong citizenship behaviours (e.g. ‘I try to represent the organisation favourably to outsiders’ – loyalty). Van Dyne et al. (1994) report Cronbach alpha coefficients of 0.83 for deviant behaviour, 0.79 for loyalty, 0.68 for social participation, 0.84 for advocacy participation, 0.75 for functional participation and 0.91 for the entire 34-item organisational citizenship behaviour scale. According to Van Dyne et al. (1994) a five-factor structure explains a large percentage (36%) of the total variance.

The Turnover Intention Scale(TIS) (Sjöberg & Sverke, 2000) was used to measure turnover intention. The scale is a five-point Likert scale, ranging from 1 (strongly disagree) to 5 (strongly agree). The scale consists of three items; an example of an item is: ‘If I was completely free to choose I would leave this job’. The Cronbach alpha coefficient for the TIS was 0.83 (Sjöberg & Sverke, 2000). Principal component analysis of the TIS in this study supported a unifactorial solution explaining 74% of the variance. The component loadings varied from 0.73 to 0.94 (Sjöberg & Sverke, 2000).

The SPSS program (SPSS, 2012) was employed to process the data. Exploratory factor analysis was performed to investigate the factor structure of measuring instruments. Descriptive statistics (e.g. means, standard deviations, skewness and kurtosis) were used to establish the normality of the distribution of scores. Cronbach alpha coefficients were used to determine the internal consistency of the four measuring instruments (Foxcroft & Roodt, 2005).

Pearson correlation coefficients were used to specify the relationships between the variables. The significance level was set at p< 0.05 (Field, 2007). Should a test statistic fall below 0.05, one can accept this as a sufficient amount of confidence to assume that the test statistic explains a great variation that is considered large enough to reflect what is genuinely taking place in the real population. Effect sizes were used to assess the practical significance of the correlation coefficients (Steyn, 2005). According to Field (2007), there are certain guidelines to assess the importance of an effect. Cut-off points of 0.30 (medium effect) and 0.50 (large effect) were set for practical significance of correlation coefficients (Field, 2007; Steyn, 2005).

A canonical correlation was used to determine the relationships between the two groups of constructs. The goal of the canonical correlation is to analyse the relationship between two sets of variables (Tabachnick & Fidell, 2001). Hierarchical multiple regression analyses were conducted with the variables in their continuous form to determine whether leader empowering behaviour and psychological empowerment predict organisational citizenship behaviour and turnover intention. A regression analysis was used in order to determine the proportion of the total variance of one variable also explained by another (Moore, 1995).

An exploratory factor analyses was conducted to investigate the factor structure of the measuring instruments. Principal component analyses were used to assess the factorability of the items on the four measuring instruments. Eigenvalues (> 1) were used to determine the number of factors in each measuring instrument. Furthermore, a principal axis factoring with a direct oblimin rotation was used in cases where scales had more than two related factors.

Principal axis factoring with a direct oblimin rotation was used on the 19 items of the LEBQ. All items on the scale produced higher than recommended communalities (> 0.30). The principal axis factoring showed that three factors could be extracted, which explained 70.89% of the total variance of the scale. The scale produced communalities that were all higher than the suggested 0.30. The first factor was labelled ‘development’ (eigenvalue = 10.49), which explained 55.26% of the total variance. The second factor was labelled ‘self-directed decision-making’ (eigenvalue = 1.94), which explained 10.23% of the total variance. Finally, the third factor was labelled ‘accountability’ (eigenvalue = 1.02), which explained a further 5.39% of the total variance. The results are in line with Van Dierendonck and Dijkstra (2012), who identified delegation of authority, accountability and facilitation (development) as the core dimensions of empowering leader behaviour.

Principal axis factoring with a direct oblimin rotation was also used on the 12 items of the MEQ. Analysis of eigenvalues (> 1) and scree plot indicated that two factors could be extracted which explains 78.16% of the total variance. The scale produced communalities that were all higher than the suggested 0.30. The first factor was labelled ‘attitude’ (eigenvalue = 7.49), which explained 62.42% of the total variance in the scale. The second factor was labelled ‘influence’ (eigenvalue = 1.89), which explained a further 13.78% of variance in the scale. Influence is comprised of items relating to self-determination and impact. This supports the results of an earlier study with service employees by Fulford and Enz (1995), who called the ‘new’ factor ’influence’. Dimitriades (2005), Fulford and Enz (1995), Hancer and George (2003) and Kraimer, Seibert and Linden (1999) suggest that self-determination and impact have something in common that is not shared with the other dimensions of the questionnaire. In their study, Hancer and George (2003) report two factors which they call ‘attitude’ (meaning and competence) and ‘influence’ (self-determination and impact).

Principal axis factoring with a varimax rotation was conducted on the 34 items of the OCBQ. Analysis of eigenvalues (> 1) and scree plot indicated that three factors could be extracted which explains 48.35% of the total variance in the scale. The scale produced communalities that were all higher than the suggested 0.30. The first factor was labelled ‘loyalty’ (eigenvalue = 12.49) and explained 35.59% of the total variance in the scale. The second factor was labelled ‘deviant behaviour’ (eigenvalue = 2.98) and explained 7.40% of the total variance in the scale. The final factor was labelled ‘participation’ (eigenvalue = 2.24), which explained 5.73% of the total variance in the scale.

A principal axis factor analysis was conducted on the three items of the TIS. Analysis of eigenvalues (> 1) and scree plot indicated that one factor could be extracted, explaining a high percentage (73.90%) of the total variance. The scale produced communalities that were all higher than the suggested 0.30. The factor extracted specified mainly as turnover intention.

Table 2 indicates that all the measures are sufficiently reliable (α > 0.60) (Nunnally, 1978), with alpha values varying from 0.61 to 0.94. Furthermore, the data on all the scales are normally distributed with skewness and kurtosis values smaller than 1.

Table 2 further indicates the correlation coefficients between the different variables. Development positively correlates practically significant (large effect) with attitude (r= 0.68; p< 0.01), influence (r= 0.56; p< 0.01), loyalty (r = 0.65; p < 0.01) and participation (r= 0.74; p < 0.01). Development has a practically significant negative relation with deviant behaviour (r= -0.53; p< 0.01) and a statistically significant negative relation with turnover intention (r = –0.21; p < 0.01).

Accountability positively correlates (practical large effect) with attitude (r = 0.73; p < 0.01), influence (r = 0.50; p < 0.01), loyalty (r= 0.69; p < 0.01) and participation (r= 0.79; p< 0.01) and negatively with deviant behaviour (r= -0.61; p < 0.01).

Self-directed decision-making positively correlates (practical large effect) with attitude (r= 0.57; p < 0.01), influence (r = 0.85; p < 0.01), loyalty (r = 0.58; p< 0.01) and participation (r = 0.56; p < 0.01). Self-directed decision-making negatively correlates statistically significantly with deviant behaviour (r= -0.23; p< 0.01) and turnover intention (r= -0.22; p< 0.01).

Attitude correlates practically significantly (large effect) with loyalty (r = 0.59; p< 0.01) and participation (r= 0.87; p < 0.01). Attitude negatively correlates practically and statistically significantly (large effect) with deviant behaviour (r = -0.50; p < 0.01). Attitude negatively correlates statistically significantly with turnover intention (r= -0.14; p < 0.01).

Influence correlates practically significantly (large effect) with loyalty (r = 0.56; p < 0.01) as well as participation (r= 0.57; p< 0.01) and statistically significantly negatively with deviant behaviour (r = -0.17; p < 0.01) and turnover intention (r= -0.21; p< 0.01).

Loyalty correlates statistically significantly with turnover intention (r= 0.23; p< 0.01), deviant behaviour correlates statistically significantly with turnover intention (r = 0.10) whilst participation negatively correlates statistically significantly with turnover intention (r= -0.14; p < 0.01).

The relationships between the different variables were tested with canonical analysis. Shown in Table 3 are correlations between the variables and canonical variates, and standardised canonical variate coefficients. Within set variance accounted for the canonical variates (% of variance), redundancies and canonical correlations between leader empowering behaviour, psychological empowerment, organisational citizenship behaviour and turnover intention set.

The results showed two statistically significant canonical variates. The Wilks-Lambada test of significance showed the results to be statistically significant (p = 0.00). According to Tabachnick and Fidell (2001), the significance should be taken at below 0.05 to indicate significance and reliability of the results. The first statistically significant canonical correlation was 0.91 with 82.92% overlapping variance. The first F-test (F_(20,631) = 33.86; p> 0.05) produced statistically significant results. The second statistically significant canonical correlation was 0.58 with 33.81% overlapping variance. The second F-test (F_(12,505) = 11.74; p> 0.05) produced statistically significant results. The third F-test (F_(6,384) = 8.08; p> 0.05) and fourth F-test (F_(2,193) = 6.84; p> 0.05) were not statistically significant. According to Tabachnick and Fidell (2001), only canonical correlations with medium to high (> 0.30) correlations should be interpreted.

The first canonical variate accounted for a more significant relationship between the different variates than the second canonical variate. Amongst the antecedent set, with a cut-off correlation of 0.30 (Tabachnick & Fidell, 2001), development (-0.83), accountability (-0.89), self-directed decision-making (-0.63), attitude (-0.59) and influence (-0.63) correlated highly with the first canonical variate. In the antecedent set, loyalty (-1.00), deviant behaviour (0.67) and participation (-0.74) correlated with the first canonical variate. Only turnover intention (0.11) in the outcome set did not correlated with the first canonical variate. This implies that the first canonical variate indicates that higher levels of loyalty and higher levels of participation, coupled with lower levels of deviant behaviour, are associated with higher levels of development, accountability, self-directed decision-making, attitude and influence. In the second canonical variate, only deviant behaviour (0.47) and turnover intention (-0.56) were significantly related. The results provide sufficient evidence in support of Hypothesis 1.

Multiple regressions were conducted to determine the predictability of independent variables on the dependant variables. Table 4 summarises the results of the regression analysis.

Table 4 indicates the results of a multiple regression with loyalty as the dependent variable and leader empowering behaviours and psychological empowerment as the independent variables. The first step of the regression analysis produced a statistically significant model (F_(3,195) = 134.22; p= 0.00), which declares 67% of the total variance. Self-directed decision-making (β = 0.35; t = 5.22; p ≤ 0.05) and accountability (β = 0.55; t = 8.58; p ≤ 0.05) made a statistically significant contribution to predicting loyalty. When psycho­logical empowerment (attitude and influence) was entered into the second step of the regression, a statistically significant model was produced (F_(5,193) = 179.44; p= 0.00). This model declared 82% of the total variance with self-directed decision-making (β = 0.20; t = 3.88; p ≤ 0.05), accountability (β = 0.27; t = 4.65; p ≤ 0.05) and attitude (β = 0.59; t = 10.04; p ≤ 0.05) significantly contributing to the variance in loyalty.

Table 5 summarises the results of the regression analysis with deviant behaviour as the dependant variable with leader empowering behaviour and psychological empowerment as the independent variables. The first step of the regression analysis produced a statistically significant model (F_(3,195) = 57.34; p= 0.00) which declares 47% of the total variance. Self-directed decision-making (β = -0.35; t = -4.11; p ≤ 0.05), accountability (β = -0.60; t = –7.41; p ≤ 0.05) and development (β = 0.40; t = 5.38; p ≤ 0.05) made a statistical significant contribution to the regression model. When psychological empowerment (attitude and influence) was entered into the second step of the regression a statistically significant model was produced (F_(5,193) = 35.14; p= 0.00). This model declared 48% of the total variance. Self-directed decision-making (β = -0.31; t = -3.49; p ≤ 0.05), accountability (β = -0.51; t = -5.23; p ≤ 0.05) and development (β = 0.31; t = 2.43; p ≤ 0.05) significantly contributed to the variance in deviant behaviour.

Table 6 summarises the results of the regression analysis with participation as the dependant variable with leader empowering behaviour and psychological empowerment as the independent variables. The first step of the regression analysis produced a statistically significant model (F_(3,195) = 66.40; p= 0.00), which declares 51% of the total variance. Moreover, self-directed decision-making (β 0.27; t = 3.23; p ≤ 0.05), accountability (β = 0.34; t = 4.35; p ≤ 0.05) and development (β = 0.18; t = 2.48; p ≤ 0.05) predict participation. When psychological empowerment (attitude and influence) was entered into the second step of the regression, a statistically significant model was produced (F_(5,193) = 44.26; p= 0.00). This model declared 53% of the total variance. Self-directed decision-making (β = -0.28; t = 3.34; p ≤ 0.05), accountability (β = 0.40; t = 4.27; p ≤ 0.05) and influence (β = 0.36; t = 3.06; p ≤ 0.05) significantly contributed to the variance in participation. Resultantly, the results provide sufficient evidence in support of Hypothesis 2.

Table 7 summarises the results of the regression analysis with turnover intention as the dependant variable and leader empowering behaviour, psychological empowerment and organisational citizenship behaviour as the independent variables. The first step of the regression analysis produced a statistically significant model (F_(3,195) = 11.36; p= 0.00), which declares 15% of the total variance. Self-directed decision-making (β = -0.39; t = -3.60; p ≤ 0.05), accountability (β = 0.48; t = 4.64; p ≤ 0.05) and development (β = -0.26; t = 2.77; p ≤ 0.05) predict turnover intention. When psychological empowerment (attitude and influence) was entered into the second step of the regression a statistically significant model was produced (F_(5,193) = 7.74; p= 0.00). This model declared 17% of the total variance. Self-directed decision-making (β = -0.33; t = -2.91; p ≤ 0.05), accountability (β = 0.62; t = 0.98; p ≤ 0.05), development (β = -0.38; t = -2.34; p ≤ 0.05) and influence (β = 0.16; t = 1.03; p ≤ 0.05) significantly contributed to the variance in turnover intention. As organisational citizenship behaviours (loyalty, deviant behaviour and participation) was entered into the third step of the regression, a statistically significant model was produced (F_(8,190) = 7.82; p= 0.00). This model declared 25% of the total variance. Self-directed decision-making (β = -0.09; t = -2.53; p ≤ 0.05), accountability (β = 0.67; t = 5.20; p ≤ 0.05), development (β = -0.16; t = -2.79; p ≤ 0.05), loyalty (β = -0.09; t = -1.77; p ≤ 0.05), deviant behaviour (β = 0.06; t = 2.78; p ≤ 0.05) and participation (β = 0.12; t = 3.58; p ≤ 0.05) significantly contributed to the variance in turnover intention. Resultantly, the results provide sufficient evidence in support of Hypothesis 3.

Pearson correlations indicated that the three components of leader empowering behaviour were practically significantly related to the components of psychological empowerment and organisational citizenship behaviour. However, deviant behaviour (as a component of organisational citizenship behaviour) negatively related practically significantly to development and accountability, and statistically significantly to self-directed decision-making. Similarly, Pearson correlations indicated practically significant relationships between the components of psychological empowerment (attitude and influence) and organisational citizenship behaviours as well as statistically significantly with turnover intention. This implies that enhancing the perceptions of psychological empowerment amongst employees will result in generating organisational citizenship behaviours amongst employees, which will decrease their intention to leave the manufacturing division.

Employees that experience their leader to invest in their development, encourage accountability, enable participative decision-making and lead by example would be more inclined to experience psychological empowerment within this context (Mendes & Stander, 2011). This would result in an increase in levels of loyalty towards the organisation and participation in organisational processes. These individuals may be less inclined to engage in deviant work behaviour. Employees will be less inclined to miss deadlines or stay away from work. Higher levels of loyalty and higher levels of participation, coupled with lower levels of deviant behaviour, are associated with higher levels of development, indicating that leaders can influence extra-role behaviour by developing direct reports. These results are consistent with various other studies examining the relationship between the variables (Chiang & Hsieh, 2012; Engelbrecht & Chamberlain, 2005).

The results denote that leadership empowerment behaviour should be encouraged throughout the organisational structures, which should result in decreased levels of turnover intention and increased levels of psychological empowerment and organisational citizenship behaviour. Within this context, managers should aim to develop and grow leadership throughout the division by incorporating empowering behaviours in their leadership styles. In support, Konczak et al. (2000) state that the leader of the future should tune into the environment by sensing the need for opportunity; engaging in kaleidoscope thinking by challenging assumptions to create new opportunities; encouraging follower buy-in by communicating inspiring visions; develop a dream by nurturing working teams; mastering perseverance and building coalitions and heroes by sharing credit and recognition with followers through the development of skills and competencies.

The results indicate that when a leader invests in the development of his or her employees, allows for decision-making and keeps employees accountable, they will experience feelings of psychological empowerment, which may lead to a willingness to be more loyal, participate in extra-role activities and be obedient to the organisation. These findings are in line with established literature (see Albrecht & Andreetta, 2010; Chen et al., 2011;Raub & Robert, 2010; Van Dijke et al., 2012; Zhang & Bartol, 2010).

Leader empowering behaviour and psychological em­powerment showed to predict organisational citizenship behaviour within this sample. Self-directed decision-making, accountability and attitude significantly contributed to the variance in loyalty. When employees feel they are in control of their work environment, they tend to find more meaning in what they do. This will result in feelings of confidence and competence, which leads to being committed and loyal towards the organisation. All the leadership empowerment constructs contributed to the variance in deviant behaviour. This implies that, when leaders empower their people (as experienced by subordinates), employees will be more willing to follow rules and procedures and produce high-quality work. Self-directed decision-making, accountability and influence significantly contributed to the variance in participation, implying that leaders who create feelings of self-determination and having an impact on the work environment (as experienced by the employees) will experience a higher level of participation from their people (Chen et al., 2011). Employees will contribute more to work-related matters, help co-workers and focus on continuous improvements (Zhang & Bartol, 2010).

Finally, a three-stage regression analysis was used to determine the predictive capacity of leadership empowering behaviour, psychological empowerment and organisational citizenship behaviour on employees’ turnover intention within the manufacturing division. Self-directed decision-making, accountability, development, loyalty, deviant behaviour and participation significantly contributed to the variance in turnover intention. The leader's ability to empower people (as experienced by subordinates) may ensure willingness of employees to become involved in in-role job performance and extra-role behaviours, preferring to stay with the organisation, even in times of difficulties. The results of this research support previous studies that indicated that a positive approach to leadership (Avey et al., 2008; Luthans, Yousef, Sweetman & Harmse, 2013; Van Dierendonck & Dijkstra, 2012; Youssef & Luthans, 2012) leads to empowerment (Avey et al., 2007; Van Dierendonck & Dijkstra, 2012), organisational citizenship (Avey et al., 2006; Chiaburu & Byrne, 2009; Chiang & Hsieh, 2012; Lavelle, 2010; Paillé, 2013) and intentions to stay within the organisation (Lam et al., 2009).

Further, the results indicate that in order for organisational talent to be retained, leadership throughout the division's structures should strive to increase employees’ feelings of self-efficacy. Employees should therefore be provided with a meaningful work experience by allowing them to gain the competence and skill they require to feel competent enough to take initiative in solving problems without supervision. By focusing on a positive approach to leadership and leadership development, organisations can contribute to employees’ experiences associated with psychological empowerment and resulting in organisational citizenship behaviours. Similarly, it could also affect the longevity of the organisation by lowering staff turnover intentions.

The use of a cross-sectional survey and the use of an availability sample merely provide a representation of employees’ opinions, thoughts, feelings and/or perceptions at one point in time. As such, causal factors associated with these subjective experiences cannot be isolated nor can their effect on the results be interpreted (Salkind, 2012). The sample size obtained limits the type of statistical analyses that could be used (e.g. structural equation modelling vs. multiple regression analysis). As a result, the current analysis procedure does not account for all the variance amongst constructs and therefore limits the extent of the research. Another limitation of the study is that the measurements were based on self-reporting methods. Self-report measures are subjective in nature and are influenced by how an employee feels at a specific point in time. These feelings might not be an accurate reflection of the entire system at the given time. The findings of the study population obtained might not be generalised because of the relatively small sample of employees in the manufacturing division.

Regardless of the study's shortcomings, the findings obtained offer valuable proactive suggestions for future research. For the study to have a greater impact, it should be stressed that the findings obtained should be replicated with larger samples forming part of the manufacturing division in order to draw more valid and reliable factor-structures for the LEBQ, MEQ and OCBQ within the South African context.

It is further recommended that a longitudinal study is conducted to establish more valid and reliable levels of leadership empowerment behaviour, psychological empowerment, organisational citizenship behaviour and turnover intention of the study population over a period of time rather than obtaining one set of perspectives, attitudes and opinions at a certain point in time, which can limit the recommendations and outcomes of the study.

Future research might explore the implications of the multidimensionality of organisational citizenship behaviour. It was found that although the three categories of organisational citizenship behaviour are correlated, employees may choose not to engage equally in all forms of citizenship behaviour. Future research should also examine when and what conditions are necessary for certain forms of organisational citizenship behaviour to occur in order to gain a better understanding of what restricts the practice of such behaviours.