Stifling My Fire: The Impact of Abusive Supervision on Employees’ Motivation and Ensuing Outcomes at Work

Introduction

Over the past two decades, much of the research on abusive leadership has focused on leaders’ behavior and its harmful effect on followers. Yet, the process by which destructive leader behavior affects followers in organizations is not fully understood (Mawritz et al., 2012). This understanding is important as this leadership style has been linked to various detrimental employee workplace outcomes. Indeed, scholars have argued that abusive supervision research lacks a unifying theoretical framework and that there are still conceptual issues that need to be addressed before the role of abusive supervision can be fully explained and understood (e.g., Mackey et al., 2017; Tepper, 2007).

Central to the abusive supervision approach is the premise that followers’ reactions to what they perceive as abusive behavior lead to negative work outcomes. Justice theory has served as a primary explanatory framework, suggesting that perceptions of abusive supervision originate from perceptions of supervisory injustice (Mackey et al., 2017; Tepper, 2007). Other perspectives emphasize the role played by retaliation towards the perceived source of abuse (Inness et al., 2005) and suggest that negative reciprocity is the driving force leading to decreased work efforts (Cropanzano & Mitchell, 2005). Although justice, retaliation, and reciprocity models have been found to provide fruitful theoretical frameworks for understanding workers’ negative reactions to abusive supervision (Cropanzano & Mitchell, 2005), they have been criticized for failing to make predictions as to when and why subordinates are motivated to perform well or engage in counterproductive behaviors (Mitchell & Ambrose, 2007). For example, Thau et al. (2007) contend that although social-exchange based explanations of employee behavior have been well supported by data, current social-exchange theories do not specify the conditions under which unjust and derogatory treatment may affect work behaviors.

We propose an alternative yet complementary outlook—one rooted not in perceptions of supervisory injustice or in reciprocity, but rather in work motivation. Drawing on self-determination theory (SDT; Deci & Ryan, 1985a, 1985b, 2000), we argue that abusive leaders affect their employees’ behaviors through their work motivation. Specifically, we explore the mediating role of subordinates’ amotivation, autonomous, and controlled motivation in the relationship between abusive leadership and employees’ job-satisfaction, counterproductive work behaviors, innovative behaviors, and turnover intentions.

Theory and Hypotheses

One of the hallmarks of leadership effectiveness is the ability to motivate followers (Kark & Van Dijk, 2007). With the introduction of SDT, motivation research shifted its focus from exploring “levels” of motivation to studying “types” of motivation and offered a distinction between autonomous and controlled motivation. One of the theory’s central tenets is that autonomy support is the most important variable contributing to motivation which in turn can lead to a variety of positive outcomes (Gagné & Deci, 2005). According to Gagne and Deci (2005), factors shown to increase autonomy support coincide with elements of communication, empathy, and concern, which are notably lacking in abusive supervision.

Drawing on self-determination theory (SDT; Deci & Ryan, 1985a, 1985b, 2000), we argue that the effect abusive leaders have on their employees occurs through the adverse impact their behaviors exert on subordinates’ autonomous motivation. We test this mechanism by focusing on employee outcomes (e.g., job-satisfaction, counterproductive work behaviors, innovative behaviors, and turnover; see Figure 1) that have far reaching impact on organizational functioning. Drawing on SDT, we expect abusive supervision to decrease employee work outcomes because it thwarts the provision of autonomy support. This is also of practical importance as leader behavior plays a key role in the development of subordinates’ motivation (Kark & Van Dijk, 2007), and individual motivation contributes substantively to organizational effectiveness at all levels including employee productivity, employee well-being, and organizational revenue (Steers et al., 2004).

Explaining Abusive Supervision Outcomes

Abusive supervision has been found to be associated with a wide range of negative organizational outcomes, such as decreased wellbeing (e.g., emotional exhaustion and job dissatisfaction; Wheeler et al., 2013; depression and anxiety; Hobman et al., 2009), decreased job performance (e.g., Xu et al., 2012), decreased organizational commitment (e.g., Duffy et al., 2002), poor interpersonal exchanges with supervisors (e.g., Lian et al., 2012), aggression (e.g., Dupre et al., 2006), and intentions to quit (e.g., Tepper, 2000). Importantly, perceptions of leadership abuse have also been found to be related to employees’ own negative behaviors at work (e.g., workplace deviance; Mitchell & Ambrose, 2007) and in their homes (e.g., work to-family conflict; Carlson et al., 2011). The longer term financial impact of negative leader behavior on organizations is noteworthy as abusive management practices have been estimated to cause businesses significant damages annually and affect many US workers (Tepper, 2007).

While in the past leadership research focused on positive leader attributes and behaviors, recently research on abusive leadership has grown as abusive leadership behaviors appear increasingly prevalent. In line with evidence of the pervasiveness and impact negative leader behaviors have on employees, numerous conceptualizations have emerged to capture this construct (Naseer et al., 2016). Defined as the “sustained display of hostile verbal and non-verbal behaviors, excluding physical contact” (Tepper, 2000, p. 178), the notion of abusive supervision captures behaviors that are particularly detrimental in that they are invisible and subjective, while far reaching, and taking place over a prolonged period of time (Hershcovis, 2011; Tepper, 2007).

A considerable body of research on abusive supervision has utilized reactance theory (e.g., Zellars et al., 2002) and an organizational justice model (e.g., Colquitt et al., 2013) to examine workers’ reactions to detrimental management practices. This approach hinges on the idea of retaliation towards the perceived source of abuse, at both the interpersonal and organizational levels (Inness et al., 2005). Unable to challenge the aggressor due to power asymmetry and lack of control over the supervisor-related stressors, ill-treated employees often react by withholding performance (Harris et al., 2007), decreasing OCBs (Zellars et al., 2002), harming the workplace (Detert et al., 2007), and exhibiting low creativity on the job (Zhang et al., 2012). Another branch of abusive supervision research draws on conservation of resources theory (Harris et al., 2007) and social exchange theory (Cropanzano & Mitchell, 2005) to explain the negative relationship between hostile management practices and subordinates’ job performance. These theoretical perspectives suggest that negative reciprocity is the driving motivation for decreased work efforts (Cropanzano & Mitchell, 2005). Simply put, mistreated individuals repay the perpetrator by diminished job performance.

In the following section, we build on previous research findings of important employee outcomes of abusive leadership. For example, Zhang et al. (2012) found that an exposure to abusive supervision can lead to a decrease in employee creativity, which is invaluable to organizational innovation and survival. Abusive supervision has also been found to adversely impact the organization through supervisor and organization directed deviance (Tepper et al., 2009; Thau et al, 2009), reduced job performance (Harris et al., 2007; Walter et al., 2015), and greater counterproductive work behaviors (CWBs; Wei & Si, 2013). Naseer et al. (2016) found that despotic leadership decreases employee in-role performance, organizational citizenship behaviors (OCBs), and creativity. We expect similar findings in our study and predict that abusive supervision will be negatively related to (a) job-satisfaction and (b) innovative behavior, and positively related to (c) CWBs and (d) turnover intentions.

Abusive Supervision and Motivation

We propose that SDT can shed light on “how” abusive supervision adversely affects employee outcomes through the type and extent of motivation employees’ experience. We argue that abusive leaders reduce employees’ autonomous motivation and increase their controlled motivation and amotivation. These motivational stances adversely affect the organization through reduced job-satisfaction and innovative behavior and increased turnover intentions and negative work behaviors (CWBs).

SDT describes a motivational continuum that ranges from amotivation (completely lacking in motivation) to intrinsic motivation (an entirely self-determined and autonomous motivation) and focuses mainly on the distinction between autonomous motivation and controlled motivation (Deci & Ryan, 1985a). A key contribution of SDT to our understanding of motivation is the ordering of extrinsic motivation along a continuum that range from lower to higher levels of self-determination (Vallerand et al., 1992). That is, while intrinsic motivation is fully autonomous, SDT provides a more fine-grained understanding of extrinsic motivation through a continuum ranging from controlled to autonomous. The recognition that there are various reasons driving individuals to engage in behaviors as a means to an end allows research to more precisely capture how various factors influence individual outcomes through their motivation. Each of the different types of motivation (e.g., amotivation, extrinsic social motivation, extrinsic identified motivation, and intrinsic motivation) conveys a different level of self-determination but is also distinct and has specific attributes that make it a category of itself. Therefore, each motivation type is best assessed as a separate construct (Gagné & Deci, 2005; Gagné et al., 2015). Intrinsic motivation, which refers to engaging with a task because it is enjoyable and interesting, represents the highest level of autonomous motivation or self-determination. Early work on SDT has used the example of “student that goes to class because he or she finds it interesting and satisfying to learn more about certain subjects” (Vallerand et al., 1992, p. 1004), a state that Albert Einstein referred to as “the enjoyment of seeing and searching” (Einstein, 1954, p. 17). Studies have shown that autonomous motivation and especially intrinsic motivation are associated with higher levels of performance and enthusiasm in various work tasks (e.g., Best, 2001) in comparison to controlled motivation or amotivation (see Gagné & Deci, 2005 for a review).

SDT portrays the impact social environments can have on motivational processes by suggesting that the social conditions in which people function can affect the extent to which they will become proactive and engaged or inactive and unenthusiastic members of a group (Ryan & Deci, 2000). Research inspired by SDT has identified these conditions across a wide range of life domains, including education (e.g., Niemiec & Ryan, 2009), parenting (e.g., Grolnick & Ryan, 1989), sports (e.g., Hagger & Chatzisarantis, 2007), health (e.g., Ryan et al., 2008), and work (Chong et al., 2020). Collectively, these studies found that situational contexts that are characterized as supportive and nurturing are conducive to proactive and engaged behaviors, and that contexts that are characterized as distressing, controlling, or threatening are conducive to behaviors that convey disengagement and withdrawal.

One important and lasting effect of leadership style is its effect on employees’ dominant motivation. As Gagné and Deci (2005) state: “people need to feel competent and autonomous to maintain their intrinsic motivation”, p. 336 ). SDT describes these needs as basic and “considered important for all individuals, so SDT research focuses not on the consequences of the strength of those needs for different individuals, but rather on the consequences of the extent to which individuals are able to satisfy the needs within social environments” (Gagné & Deci, 2005, p. 337). As a result, workplace studies demonstrate the usefulness of SDT in linking leadership styles (context) with motivation and work outcomes. For example, Chiniara and Bentein (2016) found that servant leadership, a leadership style characterized by leaders’ dedication to subordinates’ growth and empowerment, supports employees’ psychological needs, which in turn leads to task and OCB performance. Similarly, it seems possible that transformational leadership relates positively with subordinates’ autonomous motivation (Bono & Judge, 2003). While supervisor autonomy supportive behaviors, such as providing a meaningful rationale for work, acknowledging subordinates’ perspectives and contributions, and leading while offering choice rather than imposing control, have been found to increase autonomous motivation (Deci et al., 1989), a very different outcome can be expected when leadership is abusive, hostile, and derogatory (Tepper, 2000).

Research exploring the motivational mechanisms through which abusive supervision leads to negative employee outcomes is scarce, limited in scope, and thus far focused only on intrinsic motivation (Zhang et al., 2012), which represents a single polar segment of the control-to-autonomy continuum and is described as capturing only a slight portion of work motivation (Gagné & Deci, 2005). However, given that work, for the most part, may not be inherently interesting and pleasurable (i.e., intrinsically motivating), it is important to consider the entire control-to-autonomy continuum. We argue that the lack of research considering the effect abusive supervision has on varied motivational stances that have been associated with work-outcomes (e.g., autonomous and controlled motivation, and amotivation), represents a fundamental gap in this literature. For example, it is likely that abusive supervision can get employees to conform to workplace norms to avoid being punished (i.e., controlled motivation), not because they internalized the importance of work-related goals and values (i.e., identified intrinsic regulation). Moreover, the predicted impact abusive supervision has on amotivation is no less important than the impact on intrinsic motivation, given that at work one must extend effort toward organizational goal accomplishment independent of one’s level of enjoyment or engagement in the tasks.

Other studies have focused on need satisfaction as an indicator of motivation but not on actual autonomous and controlled motivation. For example, Lian et al. (2012) found that the positive relation between abusive supervision and employee deviance was moderated by employees’ basic need satisfaction. Whereas the underlying assumption in these studies is that need satisfaction increases autonomous motivation and decreases controlled motivation, a recent meta-analysis found that “satisfaction of basic needs did not substantively predict more negative forms of motivation” (Van den Broeck et al., 2016, p. 1221). Therefore, assessing positive and negative forms of motivation directly may reveal more about why abusive supervision is so damaging to individuals as leaders’ impact on subordinates’ work outcomes begins with subordinates’ motivational states. We predict that abusive supervision can be detrimental for autonomous motivation and can increase controlled motivation and amotivation.

Hypotheses 1: Abusive supervision will be (a) positively related to amotivation, (b) positively related to controlled motivation, and (c) negatively related to autonomous motivation (i.e., extrinsic identified regulation and intrinsic motivation).

Motivation and Work Outcomes

In line with extant research on SDT, research on the relations between various types of motivation and performance have found that task complexity matters. Specifically, it appears that autonomous motivation is most strongly related to performance when tasks are complex, while controlled motivation may be more positively related to performance when tasks are simple, such as in the case of assembly line work (Gagné & Deci, 2005). Additionally and in line with other motivational frameworks including Hackman and Oldham’s job characteristics model (Hackman & Oldham, 1976), it appears that tasks’ inherent interest may also affect work-related outcomes (Koestner & Losier, 2002). In spite of these differences, studies focusing on important employee outcomes, such as job satisfaction, wellbeing, and turnover, found that autonomous motivation lead to most desirable outcomes regardless of tasks complexity level (Ilardi et al., 1993; Shirom et al., 1999). As for intrinsic motivation, a recent meta-analysis considering various contingency factors supported the advantages resulting from intrinsic motivation. As the authors concluded, “it would be rare for individuals who derive personal satisfaction or enjoyment from a particular task in any context (work, school, health, etc.) to perform poorly” (Cerasoli et al., 2014, p. 996).

Research drawing on SDT found that autonomous motivation, but not controlled motivation, predict important positive work-related outcomes. For example, Chiniara and Bentein (2016) found that employees who feel they can make work-related choices, who are confident in their abilities to achieve desirable outcomes, and who feel they are cared for by and connected to others are likely to work harder and engage in OCB performance. In contrast, it appears that extrinsically controlled motivation makes individuals overemphasize attaining personal material values at the detriment of building trusting and satisfying relationships in organizations (Kasser, 2002). Prior studies have shown similar results indicating that managers’ antisocial and demotivating tactics are at the heart of employees’ reduced work commitment and CWBs (Wei & Si, 2013).

In line with previous findings indicating that when employees are treated unfairly, destructive behaviors towards the company and its members flourish (Tepper, 2007), SDT provides an explanatory mechanism toward understanding the relation between abusive supervision and subordinates’ CWBs. Mitchell and Ambrose (2007) observed a strong link between nonphysical hostile management practices and organizational and interpersonal deviance. Bennett and Robinson (2000) found that feelings of injustice, alienation, and frustration are important antecedents akin to purposeful behaviors directed against organizations, such as overstating work hours, gossiping, lateness, disobeying instructions, and theft. Finally, Lian et al. (2012) found support for a model in which abusive supervision and leader-member exchange interacted to predict employees’ organizational deviance, and with employees’ psychological need satisfaction mediating the effect of the interaction. We therefore predict that:

Hypotheses 2: The negative relation between abusive supervision and job-satisfaction will be mediated by employees’ (a) amotivation, (b) controlled motivation, and (c) autonomous motivation (extrinsic identified regulation and intrinsic motivation).

Hypotheses 3: The positive relation between abusive supervision and CWBs will be mediated by employees’ (a) amotivation, (b) controlled motivation, and (c) autonomous motivation (extrinsic identified regulation and intrinsic motivation).

Hypotheses 4: The negative relation between abusive supervision and innovative behavior will be mediated by employees’ (a) amotivation, (b) controlled motivation, and (c) autonomous motivation (extrinsic identified regulation and intrinsic motivation).

Hypotheses 5: The positive relation between abusive supervision and turnover intentions will be mediated by employees’ (a) amotivation, (b) controlled motivation, and (c) autonomous motivation (extrinsic identified regulation and intrinsic motivation).

In all mediation hypotheses, we expect abusive supervision to be negatively related with autonomous motivation, and positively related with amotivation, and controlled motivation. Then we expect autonomous motivation to be positively related with positive outcomes (i.e., job-satisfaction and innovative behavior) and negatively related with negative outcomes (i.e., turnover intentions and CWBs), and controlled motivation and amotivation to be negatively related with positive outcomes and positively related with negative outcomes.

Method

Participants and Procedure

Data was collected by Qualtrics Panel Management from a sample of 306 full-time employees working in a variety of different organizations and hierarchies in North America. All participants reported to a direct supervisor have worked under his/her supervision for three months or more, and correctly answered all careless response questions. Participants completed two online surveys 3 weeks apart to minimize the threat of common method bias as the likelihood that confounding variables will contaminate the data in a 3-week interval is relatively low (Podsakoff et al. (2003).

The majority of the respondents were female (56.5%) and their ages ranged from 19 to 68 (mean = 44.47 years, SD = 11.01). Most of the participants (52.3%) reported working with their current direct supervisor for over three years. The rest reported working with their direct supervisor between two to three years (16.3%), one to two years (14.1%), seven to twelve months (10.5%), or three to six months (6.9%). Participants’ education profile indicated the following distribution: high-school certificate (8.8%), associate degree (27.1), bachelor’s degree (14.5%), degree in medical-related fields (30.0%), master’s degree (17.0), doctoral degree (1.3%), and no certificate, diploma, or degree (1.3%).

Our participants were employed in the following industries: health care (16.3%), other services (15.4%), education (11.8%), manufacturing (9.2%), retail (9.2%), scientific and technical services (6.9%), finance and insurance (6.2%), public-administration (4.9%), construction (2.9%), trade (6%), transportation and warehousing (2.6%), and arts and entertainment (2.6%).

Measures

All measures used a 5-point scale ranging from 1 (strongly disagree) to 5 (strongly agree).

Abusive supervision was assessed at Time 1 and all other variables were assessed at Time 2. Given prior research indicating that employees can represent a valid source of performance data (e.g., Carpenter et al., 2014 found only a small mean difference between supervisor-report and self-reports of OCBs), all measures were assessed using self-reports.

“We included 4 out of the 6 components of SDT described by Gagné and Deci (2005): amotivation, controlled motivation (extrinsic social regulation), autonomous motivation (extrinsic identified regulations), and intrinsic motivation.

Abusive supervision was assessed using Tepper’s (2000) fifteen-item measure (e.g., “My boss makes negative comments about me to others”).

Work-related amotivation was assessed with a three-item measure from the Multidimensional Work Motivation Scale (Gagné et al., 2015; e.g., “I do little because I don’t think this work is worth putting efforts into”).

Work-related extrinsic social regulation was assessed with a three-item subscale from the Multidimensional Work Motivation Scale (Gagné et al., 2015; e.g., “I put effort in my job to avoid being criticized by others”).

Work-related extrinsic identified regulation was assessed with the three-item subscale from the Multidimensional Work Motivation Scale (Gagné et al., 2015; e.g., “I put effort in my job because I personally consider it important to put efforts in this job”).

Work-related intrinsic motivation was assessed with a three-item subscale from the Multidimensional Work Motivation Scale (Gagné et al., 2015; e.g., “I put effort in my job because the work I do is interesting”).

Job-satisfaction was assessed with Cammann et al.’s (1983) three-item measure (e.g., “All in all, I am satisfied with my job”).

CWBs was assessed with eight of Kelloway et al.’s (2002) ten-item measure (e.g., “I have exaggerated about my hours worked”).

Innovative behavior was assessed with Yidong and Xinxin’s (2013) three-item measure (e.g., “I develop adequate plans and schedules for the implementation of new ideas”).

Turnover intentions was assessed by reverse coding Jones’ (2010) intent to stay measure (Cropanzano et al., 1993; e.g., “The chances of me quitting my job in the next year are low”).

Results

Table 1 presents means, standard deviations, correlations, and internal consistency estimates. Following Fornell and Larcker’s (1981) recommendations, we estimated convergent and discriminant validity for our 14 variables. Convergent validity was estimated by calculating AVE (average variance extracted) values for each variable. Our findings indicate that all estimates (see CWBs as an exception) reached or exceeded the recommended value of .50 (Fornell & Larcker, 1981). The AVE value of CWBs (.46) was approaching the recommended threshold. Discriminant validity was estimated using the following procedure: for each variable we calculated the square root of the AVE. Then, we checked whether this value was larger than the correlations between the variable and other variables (Fornell & Larcker, 1981). As can be seen in Table 1, our findings indicate that all the variables met or exceeded the recommended threshold for discriminant validity.

Since we found the disattenuated estimate of the correlation between job-satisfaction and turnover intentions to be very high (r = .84; it is calculated as the division of the correlation between the two variables by the square root of the multiplication of the variables’ reliabilities), we examined an alternative model to help determine whether these two constructs are distinct enough to justify their inclusion in the study. Thus, we compared our measurement model to an alternative measurement model in which the indicators for job-satisfaction and turnover intentions load onto a single latent variable. The results of this comparison indicate that the alternative 8-factor model presents an inferior fit, χ²(820, N = 306) = 1134.159, p < .000, χ²/df = 1.383; CFI = .965, RMSEA = .035, TLI = .960, comparing to the original 9-factor model, χ²(812, N = 306) = 1076.663, p < .000, χ²/df = 1.326; CFI = .971, RMSEA = .033, TLI = .966. This comparison suggests that job-satisfaction and turnover intentions represent distinct constructs and therefore can be included in the model as separate factors.

In support of our initial prediction, abusive supervision was negatively correlated with job-satisfaction (r = -.45, p < .001) and innovative behavior (r = -.13, p < .05), and positively correlated with CWBs (r = .22, p < .001), and turnover intentions (r = .33, p < .01). We tested our hypotheses following Anderson and Gerbing’s (1988) recommended procedure for testing direct and indirect effects using structural equation modeling. First, we conducted CFA (confirmatory factor analysis), and used maximum likelihood to estimate the measurement model and then tested the structural model using AMOS 22 (Arbuckle, 2013).

We compared mean values of our focal variables across the different occupational groups in our sample by using a one-way analysis of variance (ANOVA) and found that the following 3 variables’ means are significantly different across different occupational groups: education, (F = 3.19, p = .00), age (F = 1.70, p = .04), and innovative behavior (F = 2.20, p = .00). Given these findings and since innovative behavior is a dependent variable in our study, we controlled for the effect of industry on innovative behavior in the structural model.

Measurement Model

The initial measurement model (model 1) was acceptable. However, since the significant chi-square values indicated some level of misspecification, we consulted modification indices and found that the most notable sources of misspecification were covariances between errors in CWB and in abusive leadership. After ensuring the observed variables do not load low on their respective latent factors, we linked (covariated) errors on the same factor (either CWB or abusive leadership) and found that although it did not improve the level of chi-square significance, the model fit indices improved as a result. That is, it showed that all indicators loaded highly and significantly on their respective factors, that the fit indices were all good, χ²(812, N = 306) = 1076.663, p < .000, χ²/df = 1.326; CFI = .971, RMSEA = .033, TLI = .966, and that all standardized regression coefficients were significant.

In addition to examining an 8-factor model to examine potential concerns about same-source measurement bias, we compared our 9-factor model to the following 3 models: A 5-factor model where we combined job-satisfaction with turnover intentions to a single latent factor and all motivation items to a single latent factor variable and received the following results: [χ2 (838, N = 306) = 2460.510, p < .000, χ2/df = 2.936; CFI = .819, RMSEA = .080, TLI= .796]; a 3-factor model where we combined job-satisfaction, turnover intentions, CWB, and innovative behavior into a single latent factor and all motivation items to a single latent factor and received the following results: [χ2 (848, N = 306) = 3683.689, p < .000, χ2/df = 4.344; CFI = .684, RMSEA = .105, TLI= .647]; and a 2-factor model where we combined job-satisfaction, turnover intentions, CWB, innovative behavior and all motivation variables into a single latent factor and abusive leadership items to a single latent factor and received the following results: [χ2 (850, N = 306) = 3831.468, p < .000, χ2/df = 4.508; CFI = .668, RMSEA = .107, TLI= .630]. Overall, these results indicate that the 9-factor model represents the best fit in comparison to all the other alternative models. Thus, we continued our analyses using Model 1 as our measurement model.

Common Method Variance (CMV)

To test for CMV, an additional model (model 2) was tested. In this model, we added an unmeasured methods factor to the 9-factor model and allowed all items to load on their respective factors and the methods factor. Model 2 showed a good fit, χ²(811, N = 306) = 1076.421, p < .000, χ²/df = 1.327; CFI = .970, RMSEA = .033, TLI = .965. To compare the two models (model 1 and 2), we examined CFI difference among them (Bagozzi & Yi, 1990) and found that the models were not different in their CFI values. Additionally, as suggested by Williams et al. (1989), we compared the NFI value of the measurement model with (.894) and without (.891) the method factor and found that the addition of the common factor did not improve the NFI value significantly (.003). Moreover, this value was much lower than the values reported by Williams et al. (1989) which range from .32 to .74. Finally, to estimate the magnitude of the common factor we computed the average of squared standardized loadings into the general factor in model 2 and found that the value is .047197, which indicates that the size of the common method factor is very small (Williams et al., 1989). We thus concluded that there was no indication that CMV influenced our results.

Structural Model

We examined our Hypotheses using a structural model (Figure 1). As aforementioned, since mean differences in innovative behavior were different across industries, we controlled for the effect of industry on innovative behavior in the structural model. Additionally, following the correlational findings indicating that the correlations between organizational tenure and job-satisfaction and turnover intentions, and between tenure with the supervisor and turnover intentions are statistically significant (see Table 1), we controlled for the effect of organizational tenure on job-satisfaction and on turnover intentions, as well as the effect of tenure with the supervisor on turnover intentions. We found that the results were robust to the non-use of industry as a control variable and to the non-use of tenure with the supervisor as a control variable. That is, the path coefficients between industry and innovative behavior and between tenure with the supervisor and turnover intentions came out none significant. We therefore removed industry and tenure with the supervisor as control variables in the structural model.

Our findings indicate that the model fit the data well, χ²(869, N = 306) = 1358.822, p < .000, χ²/df = 1.564; CFI = .945, RMSEA = .043, TLI = .938]. As shown in Figure 2, our findings show statistically significant path coefficients between abusive supervision and amotivation (β = .25, p < .01), extrinsic social regulation (β = .34, p < .01), extrinsic identified regulation (β = -.42, p < .01), and intrinsic motivation (β = -.29, p < .01), thus supporting Hypotheses 1a, 1b, 1c, and 1d respectively.

The Mediating Role of Motivation

According to James and Brett (1984), a complete mediation occurs when the effect of an independent variable (X) on a dependent variable (Y) disappears when a mediator (M) is added as a predictor of Y. Similarly, Baron and Kenny (1986) proposed a way to judge the effect size of an indirect effect by examining the statistical significance of the path linking X with Y (i.e., path c’). They argued that “the strongest demonstration of mediation occurs when path c’ is zero” (p. 1176). Preacher and Kelley (2011, p. 96) argue that “in practice, a researcher may claim that a mediation effect is perfect or complete if c’ is not statistically significantly different from zero” (p. 96). Therefore, as a first step, we specified a direct-effect structural model that included the independent variable abusive supervision and the four dependent variables: job-satisfaction, CWB, turnover intentions, and innovative behavior, but not the mediating variables. In this model we controlled for the effect of organizational tenure on satisfaction and turnover intentions. The model was found to fit the data well, χ²(435, N = 306) = 816.348, p < .000, χ²/df = 1.877; CFI = .940, RMSEA = .054, TLI = .927, and the path coefficients between abusive supervision and job-satisfaction (β = -.47, p < .01), CWB (β = .25, p < .01), turnover intentions (β = -.21, p < .01), and innovative behavior (β = -.15, p < .05) were all statistically significant. We then added the mediators and found that the model fit the data well, χ²(865, N = 306) = 1343.037, p < .000, χ²/df = 1.553; CFI = .947, RMSEA = .043, TLI = .939. Furthermore, the previously significant direct paths from abusive supervision to CWBs, innovative behavior, and turnover intention were no longer statistically significantly different from zero (β = -.00, p > .05; β = .05, p > .05; and β = .10, p > .05, respectively) in the presence of the mediators. This indicates that the motivation variables (i.e., amotivation, extrinsic social regulation, extrinsic identified regulation, and intrinsic motivation) may mediate the relationships between abusive supervision and CWBs, innovative behavior, and turnover intention. In contrast, the coefficient of the direct path from abusive supervision to job-satisfaction remained significant but decreased (from β = -.47, p < .01 to β = -.20, p < .01) in the presence of the mediators. Based on Baron and Kenny (1986), the fact that the direct path remained statistically significant may indicate either a partial mediation or no mediation effect. However, it is possible that the coefficient only decreased but not disappeared due to the large magnitude of c’ and not due to a lack of a mediating effect as noted by Preacher and Kelley (2011, p. 96): “In general, holding everything else constant, it is more likely that a mediator will completely mediate a relatively small total effect (c) than a relatively large total effect, so an effect in which M partially mediates a relatively large c may be more impressive than one in which M completely mediates a relatively small c.” (p. 96). To continue with the analysis, following the recommendations of Iacobucci et al. (2007) and Kline ( 2011), we removed the paths that were not statistically significant and found that the model fits the data well, χ²(868, N = 306) = 1345.788, p < .000, χ²/df = 1.550; CFI = .947, RMSEA = .042, TLI = .939. All relationships were in the hypothesized direction and statistically significant.

We then used the bootstrap procedure in AMOS 22 (Arbuckle, 2013) to estimate the significance of the indirect effects (Preacher & Hayes, 2008). We focused on direct effects and the total indirect effects of abusive supervision on CWBs (.19, p < .01; bootstrap 95% CI [.11, .38], SE = .056), innovative behavior (-.20, p < .01; bootstrap 95% CI [-.26, -.12], SE = .038), turnover intention (.25, p < .01; bootstrap 95% CI [.16, .41], SE = .062), and job-satisfaction (-.36, p < .01; bootstrap 95% CI [-.48, -.20], SE = .061). As shown in Table 2, the results indicate a partial mediation of all the four motivating variables on the relationship between abusive supervision and job-satisfaction, thus partially supporting Hypothesis 2. Moreover, our findings show that amotivation, extrinsic social regulation, extrinsic identified regulation, but not intrinsic motivation fully mediate the relationship between abusive supervision and CWBs, thus partially supporting Hypothesis 3. Furthermore, our findings show that only intrinsic motivation mediates the relationship between abusive supervision and innovative behavior, thus partially supporting Hypothesis 4. Finally, our findings show that amotivation, extrinsic social regulation, intrinsic motivation, but not extrinsic identified regulation, mediate the relationship between abusive supervision and turnover intentions, thus partially supporting Hypothesis 5.

Discussion

Our findings demonstrate that abusive supervisors’ detrimental effects on employees’ job-satisfaction, CWBs, innovative behavior, and intentions to leave the organization occur through adversely affecting subordinates’ work motivation. These results suggest that perceiving one’s supervisor as abusive is associated with approaching work with motivation that is either lacking or lower in self-determination (i.e., more amotivation and controlled motivation and less autonomous motivation). That is, it leads to experiencing work as less exciting, interesting, or fun (i.e., less intrinsically motivated), as bearing less personal significance, being less aligned with personal goals, and less important (i.e., less autonomously motivated). Additionally, such perception in turn appears to lead to experiencing work as means to avoid criticism and to gain others’ respect and approval (i.e., more control motivated), and in some cases it completely depletes employees’ motivation (amotivation). In addition, our findings indicate that abusive supervision is associated with viewing one’s work as pointless and as a waste of time, and with putting less effort at work as a result (i.e., having more amotivation). Abusive supervisors in essence lead employees to experience a negative view of work, higher intentions to quit the job, to engage less in innovative work behaviors, and to an increased tendency to engage with more deviant workplace behaviors.

Our findings are consistent with and contribute to previous research linking abusive supervision to subordinates’ functioning (Harris, 2007; Walter et al., 2015; Wei & Si, 2013). Similar to Tepper et al. (2009), we found that abusive leadership significantly increases the likelihood of workplace deviance. Our study also contributes to and extends the explanatory role of SDT in the context of employees’ attitudes and CWBs. An important tenet of SDT is that autonomy-supportive climate can enhance the development of autonomous motivation, which is seen as an important stepping stone to personal growth and lifelong development. Consistent with our hypotheses, the results of this study suggest that abusive supervision creates a work environment that is lacking in autonomy-supportive mechanisms, which are essential for employees to thrive at work. Essentially, organizations that promote or do not actively prevent such a negative environment from developing thwart their members’ development potential by not meeting their psychological needs. Our study indicates that such an environment obstructs the development of positive attitudes toward one’s job and promotes unfavorable work behaviors.

Strengths and Limitations

In expanding the nomological network of the abusive supervision construct, our study contributes noteworthy methodological strengths. The diversity of our panel member sample and several checks enforced by Qualtrics increase confidence in the generalizability of our findings.

While self-reporting of the measures was most appropriate for assessing subjective states such as autonomous and controlled motivation, job-satisfaction, and intentions to quit (Chan, 2009; Diener, 1994), the same source nature of our data remains a potential methodological limitation. Future research would benefit from including a variety of sources including third-party reports or human resources records of performance.

A second potentially important limitation is reliance on two time points. It is possible that additional observations would have yielded more accurate assessments of the relationships between our study variables. For example, employees may not engage in CWBs as a result of an isolated managerial mistreatment, but rather develop such behavior overtime as the perceived abuse repeats over a prolonged period of time (Hershcovis, 2011; Tepper, 2007).

A third limitation of our study relates to not assessing task complexity in our sample of working individuals. As aforementioned, in the context of SDT and related research, there is evidence that task complexity can act as a moderator of the relations between controlled and autonomous motivation and performance (Gagne & Deci, 2005).

Finally, a fundamental limitation of our study relates to the descriptive statistics of the abusive supervision measure we used (Tepper, 2000). We believe no psychometrical analysis has been conducted on this measure since its creation in the year 2000 although there is enough evidence to suggest that the impact of the explored phenomenon is relatively low and the samples used in many studies thus far constitute distributions that cannot be described as normal. This important caveat has been echoed by researchers including the developer of the measure himself, stating the need to improve how the abusive supervision construct is identified (Gatti et al. 2019; Tepper, 2007; Tepper et al., 2017). Nonetheless, we are not familiar with a fundamentally superior measurement of abusive supervision and, importantly, the measure we used has allowed us to better understand the construct in the context of employee motivation and work outcomes.

Future Research Directions

As our study involves a diverse sample, an important avenue for future research would be to explore contextual factors that may enhance or mitigate these effects. For example, in a context where employees are highly skilled and able to work autonomously, they may be less susceptible to have their intrinsic motivation and commitment to their goals impacted by abusive supervisors. In contrast, in contexts where employees engage daily with their supervisors and are highly dependent on them for resources for the job, it is likely that the detrimental effects of abusive supervisors on their work outcomes will be more significant.

While our data was collected over two time periods, future studies should use experimental or longitudinal designs to explore the extent to which subordinates’ autonomous and controlled motivation are affected by leadership, allowing for more definitive conclusions regarding the impact of abusive supervision on employee work motivation.

Practical Implications

Our study findings have important implications for human resource management practices. The first important step is the awareness of the pervasive motivation-reducing implications associated with hiring and promoting managers who contaminate the work environment by exhibiting such negative behaviors as rejection, lack of trust, and discouragement. This allows for a focus on also linking promotion and reward incentives of managers to morale and leadership practices that promote autonomy-supportive climate supportive of employees’ basic psychological needs for relatedness, competence, and autonomy (Ryan & Deci, 2000).

Secondly, organizations would benefit from a focus on training programs to develop manager awareness of their leadership style, as well as their leadership-related strengths and weaknesses. An important component of this training should include tools to regulate destructive behaviors and adopt positive leadership practices aimed at developing employees’ autonomous motivation.

Conclusion

Our study identifies motivation as mechanism explaining the detrimental effect of abusive supervision on subordinate workplace outcomes. Given the central importance of motivation to performance and wellbeing (Kanfer et al., 2017), our findings point to further detrimental and far-reaching effects of abusive supervisors.