The general objective of this research was to investigate the relationship between toxic leadership, job satisfaction, turnover intention and organisational commitment. The aim further was to test whether organisational culture mediates the relationship between toxic leadership and certain job outcomes, such as job satisfaction, turnover intention and organisational commitment.

Toxic leadership can and should be universally recognised as a unique set of destructive behaviours or characteristics that negatively impact the subordinate group in a predictable and intentional way (Padilla, Hogan, & Kaiser, 2007). For the purpose of this study, toxic leadership can be described as a concept that is multidimensional in nature, which compromises various dimensions as outlined by Schmidt (2008). Toxic leadership comprises five dimensions: (1) Abusive supervision refers to a leader’s perceived intentionally hostile behaviours towards employees (Dobbs, 2014; Schmidt, 2014), excluding physical contact (Tepper, 2007); (2) Authoritarian leadership attempts to exert excessive authority and control over subordinates (Dobbs, 2014) in such a way that the leaders ultimately control all the work (Schmidt, 2014); (3) Narcissism points to a style driven by arrogance and self-absorption, where self-orientated actions are designed to enhance the self (Dobbs, 2014) but often fails to follow company policies whilst expecting it of employees (Schmidt, 2014); (4) Self-promotion advertises their accomplishments and also take credit for others’ work (Dobbs, 2014), blame others and deflect responsibility for mistakes (Schmidt, 2014); and (5) Unpredictability: through their actions, they keep subordinates afraid and watchful (Dobbs, 2014). These leaders act differently when their superiors are around (Schmidt, 2008), and their consistently unpredictable actions eventually cause their subordinates to give up, feeling helpless and powerless to protect themselves (Schmidt, 2014).

The study followed a quantitative research approach and used a cross-sectional research design. This form of research focusses on gathering numerical data and generalising it across groups of people or explaining a particular phenomenon (Babbie, 2010). The term ‘cross-sectional’ in a research design refers to the collection of data in more than one case, and at a single point in time. For this study, the advantage was that the data could be collected at a single point in time and from a specific population in the manufacturing sector (Struwig & Stead, 2013).

The Human Resource Managers of the various companies acted only as gatekeepers, ensuring access to the participants to ensure data collection. In total, 800 questionnaires were distributed to the various organisations, and 600 were collected. Data collection was done in the manufacturing sector and in the work environment, which is familiar to the researcher, using a questionnaire booklet that was individually administered. Accompanying the questionnaire booklets was a letter of introduction and an explanation of the objectives and importance of the study. Participants were accommodated in the training centre or empty office space of the organisations to complete the questionnaires.

Although this was a self-administrated questionnaire, a field worker working as a human resource intern was available at one of the organisations to assist participants who found it difficult to complete the questionnaire because of low literacy levels. The fieldworker was trained to assist the participants during the data collection phase. No personal information was recorded on the booklet, to ensure anonymity, respect privacy and ensure confidentiality (De Vos et al., 2011). Furthermore, no ethical guidelines were breached by the researcher during the process. The researcher maintained the respect and dignity of the participants throughout and adhered to the ethical considerations (De Vos et al., 2011). Sufficient time was allocated for the participants to complete the survey.

The researcher coordinated the distribution of the questionnaires, as well as the collection and safe storing of completed booklets. After all the booklets were collected, the data were captured and the statistical analysis commenced.

Statistical analysis was carried out using the SSPS program version 25 (IBM SPSS Inc., 2017) and AMOS 20 (Arbuckle, 2011). Data were analysed through descriptive and inferential statistics such as means, standard deviation, skewness and kurtosis. The Cronbach’s alpha coefficients were calculated to determine reliability, with values of 0.70, which according to Pallant (2010), can be regarded as reliable. Confirmatory factor analysis (CFA) was done on the different scales of toxic leadership, turnover intention, organisational commitment, job satisfaction and organisational culture. The results for all scales indicated that they could be used in their present form to complete the statistical analysis.

To determine the relationships between variables (i.e. toxic leadership, job satisfaction, organisational commitment and turnover intention) and the strength of these relationships, product–moment correlations and Pearson’s correlation coefficient were used. The correlation coefficient cut-off scores were set at −1 to +1, with −1 indicating a negative relationship, 0 no relationship and +1 a positive relationship. Statistical significance was set at 95% (p ≤ 0.05). Practical significance was determined using effect size, and its cut-off point was set at 0.30 for medium effect and 0.50 for large effect (Steyn & Swanepoel, 2008).

The aim was to test whether organisational culture has a mediational effect between toxic leadership and certain job outcomes such as job satisfaction, commitment and turnover intention. For this test, structural equation modelling (SEM) was used as implemented by AMOS (Arbuckle, 2011). Several goodness-of-fit indices were used, namely, chi-square (χ²), the Comparative Fit Index (CFI), Tucker-Lewis Index (TLI) and the root mean square error of approximation (RMSEA). Acceptable goodness-of-fit indices had the following cut-off points: non-significant χ² values; CFI and TLI values larger than or equal to 0.90; and RMSEA values smaller than or equal to 0.08 (Byrne, 2010).

It is evident from the results that most variables were found to be distributed normally. The measuring instruments also showed acceptable levels of internal consistency. Furthermore, the results also indicate Cronbach’s alpha coefficients with acceptable levels of reliability. These coefficients range between 0.66 and 0.85. All the Cronbach’s alpha values were found to be above the cut-off point of α ˃ 0.70, except for normative commitment, cultural management, performance orientation and task structure, but these are still regarded as reliable (Pacleb & Cabanda, 2014; Pallant, 2010; Pevalin & Robson, 2009).

The correlation coefficients between the constructs are presented in Table 2.

The results of the mediation analysis are discussed and displayed in Figure 1.

Structural equation modelling was used to test the proposed mediating effect of organisational culture between toxic leadership and job satisfaction, organisational commitment and turnover intention. The hypothesised mediation model above tested whether organisational culture mediates the relationship between toxic leadership and job outcomes, namely, turnover intention, job satisfaction and organisational culture. The results indicated a significant negative relationship between toxic leadership and organisational culture (R² = −0.50; SE 0.05; p = 0.00). Furthermore, a significant negative relationship was found between organisational culture and job satisfaction (R² = −0.48; SE 0.02, p = 0.00). The results indicated a significant positive relationship between organisational culture and organisational commitment (R² = 0.56; SE 0.04, p = 0.00), and a significant negative relationship between organisational culture and turnover intention (R² = −0.51; SE 0.05, p = 0.00).

The overall model was tested in three separate models, the results of which are displayed below. The mediation was partial for all the models that were tested and explain 45.8% of the variance in job satisfaction, 63.5% of the variation in organisational commitment and 71.6% of the variance in turnover intention.

Table 3 displays the goodness-of-fit statistics for the three models.

Marsh, Hau and Wen (2004) cautioned that an acceptable model can be rejected if there is a strict adherence to cut-off values. Marsh et al. referred to a traditional cut-off values amounting to “little more than rules of thumb based largely on intuition and have little theoretical justification” (2004, p. 321). Cut-off points should only be considered as guidelines, as there is little consensus on the values for adequate fit (Lance, Butts, & Michels, 2006). According to Hooper, Coughlan and Mullen (2008), researchers must not lose sight of the substantive theory. In other words, by allowing model fit to drive the research, it moves the research away from the original theory-testing purpose.

As mentioned above, the overall model was tested in three separate models and is described below.

Model 1 tested the proposed mediating effect of organisational culture between toxic leadership and job satisfaction. The hypothesised models all provided an adequate fit to the data. The CFI value for Model 1 was found to be at 0.90, the TLI (0.88) value just below the cut-off point of 0.90 and RMSEA was 0.09, which is indicative of a fit that is neither good nor bad. Model 1, therefore, shows acceptable model fit.

Model 2 tested the proposed mediating effect of organisational culture between toxic leadership and commitment. In terms of Model 2, both the CFI (0.89) and TLI (0.87) values were found to be just below the cut-off point, but RMSEA was 0.09, which is indicative of a fit that is neither good nor bad. Model 2 shows weak but acceptable fit.

Model 3 tested the proposed mediating effect of organisational culture between toxic leadership and turnover intention. Model 3 has TLI (0.86) and CFI (0.88) just below the acceptable cut-off point. Root mean square error of approximation was 0.09, which indicates a fit that is neither good nor bad. Therefore, Model 3 shows weak but acceptable fit.

Table 4 displays the results of the mediating effects of organisational culture. The mediation model tested whether organisational culture mediates the relationship between toxic leadership, job satisfaction, organisational commitment and turnover intention. The model does not specify whether the mediation is full or partial. The test results returned both direct and indirect effects for all the relationships tested, which indicated that mediation is only partial in all of the relationships. In terms of job satisfaction, results indicated that organisational culture does mediate the relationship between toxic leadership and job satisfaction. This mediation is found to be partial and explained 45.8% of the variance in job satisfaction. In addition, organisational culture also mediates the relationship between toxic leadership and organisational commitment. This mediation is partial and explains 63.5% of the variation in organisational commitment. In terms of the results, it confirmed that organisational culture does mediate the relationship between toxic leadership and turnover intention. This mediation was found to be partial and explains 71.6% of the variance in turnover intention.

According to Mehta and Maheshwari (2014), the behaviour and performance of leaders must be monitored and assessed constantly to ensure their interactions with staff are conducive to healthy work environments. Organisations often lack the experience and abilities to counteract the effects of toxic leadership, thus usually being forced to pay the hidden costs inflicted by the dysfunctional behaviour of the toxic leader. These hidden costs may entail: reduced productivity because of an increase in absenteeism and sick leaves; weak performance of employees caused by commitment and dissatisfaction at work; decreased brand equity as a result of reputational damages of the organisation and legal costs (Vreja, Balan, & Bosca, 2016). It is anticipated that studies such as these will encourage the organisations to focus more on instances of toxic leadership within and help them put checks and balances in place to ensure the early detection of toxic leaders (Mehta & Maheshwari, 2014). In this regard, this study will help initiate interventions, where management within organisations can focus on dealing with identified toxic leaders.

Leadership studies have indicated that employees tend to blame the organisation as a whole for having a culture that tolerates toxic leadership. Employees respond to this perceived culture of tolerance by being negatively inclined towards the organisation as a whole (Folger & Cropanzano, 2001). This study has indicated that such a negative response among employees is true to an extent. It has provided empirical data to support the statement that toxic leadership has a significant impact on job outcomes, and in turn on employees having to cope with such an organisational culture.

The focus of the research, being a novel topic in the South African context, has expanded on the knowledge of toxic leadership. In this regard, this study as such has increased the opportunity to deal effectively with the destructive impact of toxic leadership on employees and organisations at large.

A practical implication for employees in the manufacturing industry is that this study provides them with an explanation of toxic leadership as a leadership style, and how it is intertwined with an organisation’s culture. The results explain further as to how this leadership style may affect employees in the workplace: an insight which could help them make crucial decisions about their jobs and future career. This study provides human resource professionals an in-depth understanding of the phenomenon of toxic leadership. This focus may be a starting point to help organisations decrease the prevalence of toxic leadership styles, thereby reducing its destructive impact within the workplace.

Firstly, the study was confined to the manufacturing sector in South Africa. As indicated, such a restriction of the study population and the sampling procedure may influence the generalisability of the results. Future research should, therefore, replicate the study in different sectors that represent different populations. Such a broadened research scope would provide a true reflection of toxic leadership in South Africa and indicate whether the results in the various sectors are similar or differ considerably.

Secondly, whilst self-promotion had the strongest impact on job outcomes, certain other dimensions showed an amount of impact as well. Thus, it would be advantageous to organisations if research could investigate the variance that is unique to all five dimensions of toxic leadership.

Thirdly, this study confirmed relationships as well as predictors in a small section of the manufacturing industry. Therefore, it would be valuable to continue these studies in other manufacturing organisations to generate more generalised results. Such a focus applies especially to this study’s unique finding on the positive experience of job satisfaction. This finding is not in line with other studies; therefore, future research should replicate the study to assess whether this result is specific to this study population only.

In addition, the study should be expanded into other sectors to confirm the relationships between toxic leadership and job outcomes, predictors and organisational culture as a mediator. Such future research would be valuable, especially in light of the fact that currently there are no other studies in South Africa with which results can be compared. It would also be recommended that this study should compare if the same results occur in various occupation levels from the same population.

Finally, although the mediation results indicated directional relationships between toxic leadership and job outcomes, the data were not collected over an extended period of time; consequently, it was not possible to test the assumptions empirically. Therefore, the use of longitudinal studies can assist by measuring directional relationships over an extended period of time. Such an approach will deliver richer data and have positive implications for operational interventions, because it may give a deeper insight to the cause and effect aspects (De Vos et al., 2011).

This study provided the first analysis of toxic leadership within a South African working environment. The findings indicated the impact of this leadership style on certain job outcomes such as turnover intention, job satisfaction and organisational commitment. The study also investigated the mediation role of organisational culture. It is important that other researchers build on the above-mentioned recommendations for future studies, because several questions remain unanswered, for example, the unexpected difference in the result of job satisfaction.