Abstract
Although coping with an abusive boss can be psychologically demanding, those who suffer from leader abuse often stay in these unpleasant relationships, actively managing the way they are viewed in the eyes of their abusive leader (source of the abuse) and coworkers (observers of the abuse). Accordingly, the abusive supervision literature has relied almost exclusively on an emotional appraisal perspective to study the self-image implications following leader abuse. The present study seeks to add to this emerging line of scholarly conversations by presenting a novel theoretical alternative. Specifically, we integrate self-presentation and self-concept orientation perspectives to portray individuals’ identity-driven self-image maintenance following leader abuse. We argue that only those with a stronger relational self-concept are likely to be motivated to preserve their identity-consistent self-image and present themselves in positive and socially desirable ways toward both their coworkers and leader, following leader abuse. Using survey data collected from working professionals in China across two field studies, we found support for our hypotheses that when employees with a stronger relational self-concept experienced abusive supervision, they were motivated to help their coworkers as a result of their relational reputation maintenance concerns, and to use ingratiation tactics toward their leader due to their image preservation motives. We also offer insights about both the theoretical and practical implications of our research and discuss study limitations and directions for future research.
Mounting evidence shows that abusive supervision, a leader's “sustained displays of hostile verbal and nonverbal behaviors” (Tepper, 2000: 178), imposes significant adverse impacts on employees, resulting in dysfunctional work attitudes and counterproductive behaviors (e.g., Mawritz, Dust, & Resick, 2014; Tepper, Simon, & Park, 2017; Whitman, Halbesleben, & Holmes, 2014). Nevertheless, those who suffer under an abusive leader quite often feel compelled to maintain the unpleasant relationship (Tepper, Moss, Lockhart, & Carr, 2007). Instead of responding with hostility or retaliation (Lian et al., 2014a; Tepper, Henle, Lambert, Giacalone, & Duffy, 2008), some mistreated employees may blame themselves for leader abuse and engage in compensatory prosocial behaviors (Yu & Duffy, 2021) to actively manage the way they are perceived in the eyes of the leader (Vogel & Mitchell, 2017). This shift of focus from victims’ dysfunctional social exchange to their concern for self-image has predominantly relied on the social functional view of emotions and focused on victims’ emotional appraisals (e.g., feelings of anger, shame, or guilt) following leader abuse (Peng, Schaubroeck, Chong, & Li, 2019; Troester & Van Quaquebeke, 2021; Yu & Duffy, 2021).
The fruitful findings from these recent studies, however, point to the need to better understand the conditions under which subordinates’ experience of leader abuse triggers their self-image concerns. Specifically, victims’ concern for self-image has been found to be largely dependent on contingent factors such as social comparisons of leaders’ treatment of subordinates (Peng et al., 2019), the quality of the dyadic social exchange relationship (Troester & Van Quaquebeke, 2021), or their attribution of leadership motives (Yu & Duffy, 2021). An intriguing question thus remains with respect to the stable characteristics that may account for individual variation in victims’ motivation for self-image maintenance (Troester & Van Quaquebeke, 2021; Vogel & Mitchell, 2017; Yu & Duffy, 2021). This is especially important to study as it helps explain why some individuals can effectively manage the interpersonal dynamics in the context of abusive supervision while some cannot (Tepper et al., 2017).
In the present study, we enrich this line of research by integrating self-presentation (Leary & Kowalski, 1990; Schlenker & Leary, 1982) and self-concept orientation (Brewer & Gardner, 1996) perspectives to explain followers’ identity-driven self-image maintenance following leader abuse. According to self-presentation perspectives, individuals have a pervasive desire to be viewed favorably and thus are motivated to influence how they are perceived by others in the social environment (Jones & Pittman, 1982; Schlenker, 1980) especially when their motivation for self-image maintenance is rooted in their deeply held belief structures (Leary & Kowalski, 1990; Schlenker & Weigold, 1992). Given that leader abuse indicates a challenge to one's relational value (Tepper, 2000, 2007), we argue that experiencing abusive supervision creates self-image concerns particularly for those with a stronger “relational” self-concept, defined as a self-concept orientation built on connections to other people through interpersonal ties and quality relationships (Brewer & Gardner, 1996). These individuals tend to value interpersonal bonding and harmonious relationships and are sensitive to signals of hostility, public insult, or social exclusion (Schilpzand & Huang, 2018).
To further illustrate identity-consistent self-image maintenance mechanisms following leader abuse, we build on one fundamental premise in the self-presentation literature: People tend to use different self-presentation strategies as they engage different audiences (Baumeister & Tice, 1984; Leary et al., 1994; Tice, Butler, Muraven, & Stillwell, 1995). Differentiating between the unique paths and outcomes representing victims’ self-image maintenance effort toward different targets is necessary as self-presentation is a goal-directed process with specific objectives targeting a particular audience (Leary & Kowalski, 1990). Thus, we expect that abused employees with a stronger relational self-concept will choose to maintain a desirable identity-consistent self-image in distinct ways and for different reasons, as they interact with their coworkers, who are observers of the abuse (Mitchell, Vogel, & Folger, 2015; Priesemuth & Schminke, 2019), and their leader, who is the source of the abuse (Tepper et al., 2007).
Specifically, we expect that abused employees with a stronger relational self-concept will develop relational reputation maintenance concerns (relational RMC), which we define as an individual’s desire to retain a positive image as a person who is known for fostering and maintaining harmonious interpersonal relationships. Relational RMC reflects a concern, rooted in self-presentation motives, for acceptance from others in the broader social environment (De Cremer & Tyler, 2005). Thus, relational RMC should further motivate victims of leader abuse to craft an identity-consistent image in the work group by engaging in impression manipulating behaviors that are prosocial in nature. Particularly relevant in this context is helping behavior toward coworkers (i.e., the observers of leader abuse), defined as giving assistance in order to facilitate the work group's task accomplishment (Van Dyne & LePine, 1998). Giving help is not only perceived positively by the beneficiary (i.e., coworkers), but also carries interpersonal currency and creates interdependence between the two parties as it signals that those who give help are relationally valuable and can be counted on (Carnevale, Huang, Vincent, Farmer, & Wang, 2021; Grodal, Nelson, & Siino, 2015), despite being mistreated by the leader. This is consistent with prior research which suggests that individuals may use helping behavior in order to solicit favorable impressions from others (Bolino, 1999; Grant & Mayer, 2009).
Similarly, we expect that victims with a stronger relational self-concept are also motivated to attenuate the self-image deficiency as they face the source of abuse (their leader). We describe this underlying mechanism as “image preservation motives,” defined as a person’s desire to preserve a favorable impression in the eyes of others (Ng & Yam, 2019; Tetlock & Manstead, 1985). Image preservation motives are self-enhancing in nature and tend to promote behaviors that help improve one's social image (Ng & Yam, 2019). We position image preservation motives in the context of leader-follower interactions to capture victims’ desires to actively influence how their leader views them so as to break the cycle of leader abuse. These image preservation motives should lead them to engage in ingratiating behavior toward the leader, which has been found to be an effective self-presentation tactic directed toward the leader (Koopman, Matta, Scott, & Conlon, 2015). We thus consider ingratiation to be a unique outcome of image preservation motives because it is other-enhancing and conforming in nature (Bolino & Turnley, 2003; Lee, Quigley, Nesler, Corbett, & Tedeschi, 1999), fostering a more favorable image with the abusive leader via impression-manipulating behaviors such as flattery or opinion conformity.
Our study makes three important contributions to the extant literature. First, while recent research has approached self-appraisal following leader abuse almost exclusively from the social functional view of emotions (Peng et al., 2019; Troester & Van Quaquebeke, 2021; Yu & Duffy, 2021), we add to this line of research by integrating self-presentation and self-concept orientation perspectives to portray self-appraisal processes following leader abuse as identity-driven self-image maintenance. Second, answering the call from recent studies (Vogel & Mitchell, 2017; Yu & Duffy, 2021), we take a pioneering role in exploring individual characteristics that can explain why certain people are more likely than others to sense and act upon a need for self-image maintenance following leader abuse. Finally, while recent research shows the potential for socially beneficial reactions from victims of leader abuse toward either the observer (i.e., coworkers) or the source (i.e., leader) of abuse (Troester & Van Quaquebeke, 2021; Yu & Duffy, 2021), we add to this emerging line of research by illuminating the dual mediating mechanisms through which the abused employees maintain their self-image toward both the observer and the source of the abuse simultaneously, and thus also provide further empirical evidence of victims’ self-presentational reactions following leader abuse. We present our model in Figure 1.
Proposed Research Model of Identity-Consistent Self-Image Maintenance Following Leader Abuse
Motivation for Goal-Directed Self-Image Maintenance
Individuals differ in their image-crafting goals as they engage different audiences, which is reflected in their choice of self-presentation tactics (Baumeister & Tice, 1984; Leary et al., 1994). Thus, it is crucial to understand the nature of victims’ goal-directed efforts toward different audiences in the context of abusive supervision—the observer (i.e., coworkers) of the abuse and the source (i.e., the leader) of the abuse. Regarding self-image maintenance toward coworkers, we focus on relational RMC, which reflects individuals’ cognitive appraisals of potential loss to their image as a person who can effectively manage relationships. As such it represents a specific type of RMC, which, in its more general form, has been described as individuals’ desires to retain the positive image others have of them (Baer et al., 2015). According to Baer et al. (2015), individuals consider their reputation in a particular area as a valuable resource and thus perceive challenges to their reputation as a salient concern. A person’s reputation is different from their workplace status—which refers to a more general assessment of their overall respect, prominence, and prestige (Djurdjevic et al., 2017)—and instead describes their projected image toward a specific domain (Ferris, Blass, Douglas, Kolodinsky, & Treadway, 2003; Zinko, Ferris, Humphrey, Meyer, & Aime, 2012). As described in Klotz and Bolino (2013), reputation is built from an evaluation of patterns of specific past behaviors, based on which expectations are created for future actions. Recent research has explored specific domains of RMC, such as creative RMC, or concerns for preserving a reputation for creativity (Carnevale et al., 2021). In the present study, we position relational RMC in the broader social environment where image crafting matters for acceptance by others (De Cremer & Tyler, 2005) due to the presence of an abusive leader. That is, relational RMC is important in the context of victim-observer dynamics as it reflects a victim’s goal (and the associated heightened psychological pressure) to prove their value in a social environment, particularly when the victim cares deeply about defending their self-view as someone who desires to maintain quality interpersonal connections and social relationships (i.e., people with a stronger relational self-concept).
Regarding self-image maintenance toward the leader, we focus on victims’ image preservation motives. Preserving a positive self-image not only indicates one's desire to avoid potentially negative evaluations, but also includes one's motivation to create conditions that can facilitate more favorable impressions from the target audience (Jones & Pittman, 1982). This self-enhancing nature of image preservation motives is particularly evident in interpersonal interactions where actors need to defend their self-esteem (Tedeschi & Melburg, 1984). Early contributions to the self-presentation literature have also suggested that individuals often present themselves in a way that conforms to the preferences of the target audience if they are somehow dependent on the target for desired outcomes (Jones & Pittman, 1982; Schlenker, 1980). Conforming to the target audience becomes particularly important when the actor is subordinate to the target (Schlenker & Weigold, 1992). In the context of abusive supervision, recent research suggests that victims of leader abuse often make deliberate efforts to “break the cycle” of leader abuse (Wee, Liao, Liu, & Liu, 2017) or to actively use self-presentation tactics to regain their diminished self-esteem (Vogel & Mitchell, 2017), thus indicating an image-preservation motive for the victim—particularly those with a stronger relational self-concept—to present themselves toward the leader in a way that could help to attenuate further leader abuse.
In the present study, we expect that a person’s experience of leader abuse alone would not necessarily result in their motivation for self-image maintenance. Instead, we argue that one's identity orientation (i.e., stronger relational self-concept) serves as a catalyst for victims of leader abuse to feel the need for self-image protection and, subsequently, to choose to act upon this need in a way consistent with their deeply held belief structures. This is consistent with the self-presentation literature which suggests that when individuals are confronted with negative evaluations, only those who believe such evaluation contradicts how they view themselves will be more sensitive to such an assessment and thus choose to pursue self-enhancing goals, that is, improving their self-image via presenting themselves more strongly in a way that is consistent with how they view themselves (Schlenker & Weigold, 1992). Those with a weaker relational self-concept, on the contrary, might care less about the potentially damaging effect of a dysfunctional social relationship on how they are viewed in the eyes of others or the leader. Therefore, they are less likely to actively maintain a favorable image as perceived by their coworkers or their leader. That is, we expect that a person’s experience of abusive supervision would only trigger their self-image maintenance if they have a strong relational self-concept.
Identity-Driven Self-Image Maintenance Toward Coworkers
Abusive supervision is associated with a variety of adverse employee outcomes (Greenbaum, Hill, Mawritz, & Quade, 2017; Lian, Ferris, & Brown, 2012; Lian et al., 2014a; Lian, Ferris, Morrison, & Brown, 2014b; Mitchell & Ambrose, 2007; Tepper et al., 2007). While these prior studies drew primarily from either the (in)justice or social exchange perspective (see Mackey, Frieder, Brees, & Martinko, 2017, for a review), recent studies have relied on an emotional appraisal lens and have found that leader mistreatment can have negative implications for victims’ self-image (Oh & Farh, 2017; Peng et al., 2019; Yu & Duffy, 2021). For instance, Yu and Duffy (2021) found that abused employees experienced guilt when they believed leader abuse was a result of a performance deficiency, leading them to craft a more favorable social image as a good organizational citizen. Troester and Van Quaquebeke (2021) further showed that a victim’s impaired self-image could motivate them to help their abusive boss.
In the present study, we break new theoretical ground by drawing on the self-concept orientation perspective to account for an identity-driven explanation to describe self-image maintenance following leader mistreatment. Per Brewer and Gardner (1996), the self-concept reflects three levels of self: personal, relational, and collective. The relational self-concept—which is of particular interest in this study—refers to an identity orientation where the self is primarily defined by relationships with others rather than as a separate, autonomous individual (personal), or through memberships in large groups (collective; Cooper & Thatcher, 2010). Each person has all three levels represented in their self-concept, but individuals differ in how important each is to them (Brewer & Chen, 2007). We expect that only victims with a stronger relational self-concept are likely to experience relational RMC toward coworkers because these individuals are not only sensitive to signals of hostility or insult but are also motivated by their identity-driven relational goals (i.e., maintaining harmonious social relationships with others) to manage situations where their socially desirable image, in the eyes of others, is challenged by leader mistreatment.
Recent studies show that victims of leader abuse may deal with significant social consequences (e.g., being socially ostracized) from coworkers who blame (Mitchell et al., 2015) or have a rivalry with the victim (Xu et al., 2020), suggesting that experiencing leader abuse can signal an incapacity to effectively manage relationships. Victims of abusive supervision experience concern about how they manage their relationship with their leader and therefore engage in more upward maintenance communication as a way to improve a dysfunctional leader-subordinate relationship (Tepper et al., 2007).
As described in the self-presentation literature (Bozeman & Kacmar, 1997; Leary & Kowalski, 1990), individuals experience a psychological state that drives their self-image protection when their self-image is challenged. As such, concern for being seen as relationally deficient as a result of experiencing leader mistreatment should be salient primarily for individuals who have a stronger relational self-concept because they value interpersonal bonding and connection to others in the social environment (Brewer & Gardner, 1996; Cooper and Thatcher, 2010). Indeed, individuals’ concern for maintaining a socially desirable self-image may trigger their heightened feeling of psychological pressure when their motivation for self-image maintenance is derived from their deeply held values or belief structures (Leary & Kowalski, 1990; Schlenker & Weigold, 1992). Prior research also found that those who cared about their public appearance felt compelled to maintain a socially desirable self-image particularly when doing so would align with their social identity (Schlenker & Weigold, 1990). Following such logic, we argue that a person's identity orientation functions in a way that determines their need for self-image maintenance. Therefore, when a favorable self-image is challenged, individuals experience concern and a desire to protect their identity-consistent self-image. As such, concern for being seen as relationally deficient or dysfunctional, as a result of experiencing leader mistreatment, should be salient only for individuals with a stronger relational self-concept. In sum, we expect that only for those with a stronger relational self-concept will the experience of leader abuse motivate them to keep intact their social image of managing social relationships well. We thus hypothesize the following:
Hypothesis 1: Abusive supervision and relational self-concept will interact to predict relational reputation maintenance concerns, such that abusive supervision will have a significant positive effect on relational reputation maintenance concerns only in the presence of a stronger (vs. weaker) relational self-concept.
We further propose that victims with a stronger relational self-concept will, when prompted by their relational RMC, present themselves as socially desirable. This self-image maintenance is necessary because, although the abuse itself takes place in a dyadic relationship, most often the abusive relationship and its consequences are known in the broader social context (Priesemuth, Schminke, Ambrose, & Folger, 2014; Schaubroeck, Peng, & Hannah, 2016). Therefore, the self-presentation perspective (Schlenker & Leary, 1982; Leary & Kowalski, 1990) would predict that those with a stronger relational self-concept will feel increased psychological pressure to engage in behaviors that can help restore their damaged social image. This is consistent with recent findings about the motivational role of domain-specific RMC in encouraging individuals to be more agentic in their effort to defend and promote their reputation (Carnevale et al., 2021).
In the present study, we focus on helping behavior—which has been conceptualized and operationalized as a form of organizational citizenship behavior (OCB; Van Dyne & LePine, 1998)—as a particularly relevant self-image maintenance outcome because it benefits coworkers and is likely to affirm a helper’s commitment to coworker relationships. Indeed, helping behavior has been described as an effective interpersonal tactic to seek favorable impressions from others in the workplace (Bolino, 1999; Grant & Mayer, 2009). It can be a particularly attractive impression manipulation strategy when the actor sees it as instrumental to goal accomplishment and desires to acquire intangible resources, such as a reputation for doing so (Hui, Lam, & Law, 2000). Therefore, when confronted with a potential challenge to their reputation as someone who values and is capable of managing social relationships, victims with a stronger relational self-concept should be motivated to engage in activities directed toward improving their social image in the eyes of the observers (Jones & Pittman, 1982; Tedeschi & Melburg, 1984). That is, the self-presentation perspective (Leary & Kowalski, 1990; Schlenker & Leary, 1982) would suggest that those holding a stronger relational self-concept when experiencing leader abuse, should be motivated to help their coworkers because engaging in helping OCB can help restore their favorable relational image, proving that they can still provide relational value in the broader social environment. On the contrary, we expect that those with a weaker relational self-concept would be unlikely to engage in helping behavior toward their coworkers because restoring and maintaining a favorable social image is not important to their identity beliefs. We thus hypothesize the following:
Hypothesis 2: Abusive supervision will have a positive indirect effect on employees’ helping behavior via relational reputation maintenance concerns only in the presence of a stronger (vs. weaker) relational self-concept.
Identity-Driven Self-Image Maintenance Toward the Leader
When deeply held identity beliefs are challenged, individuals are motivated to rely on the use of self-presentation tactics to attenuate their self-image deficiency (Leary, 1994; Schlenker & Weigold, 1990). Recent studies have offered preliminary evidence showing that, instead of fighting back against leader abuse with negative or aggressive behaviors (Lian et al., 2014a, 2014b), abused employees might choose to either decrease their resource dependence on the abusive leader (Wee et al., 2017) or even respond to leader hostility with a “warm heart” by helping their abusive boss (Troester & Van Quaquebeke, 2021). Here, our focus is on how such responses may be rooted in efforts to bolster their self-concept orientations.
Individuals consistently consider how others view them and work to build a self-image in line with their identity beliefs (Leary & Kowalski, 1990). Such concern for self-image has been evidenced in recent abusive supervision research which drew primarily from an attributional view of self-blame for leader mistreatment (e.g., Troester & Van Quaquebeke, 2021; Yu & Duffy, 2021). Although grounded in different theoretical roots, these studies underscore the concern for self-image that abused employees may develop. Integrating self-presentation (Leary & Kowalski, 1990; Schlenker & Leary, 1982) and self-concept orientation (Brewer & Gardner, 1996) perspectives, we focus on victims’ identity-based appraisal of their experienced mistreatment and their subsequent desire to maintain a favorable self-image in front of the abusive boss. Employees with a strong relational self-concept will be concerned about preserving their relationally-oriented identity beliefs and are more likely to engage in behaviors that help eliminate disapproval from the abusive leader. Moreover, the self-presentation literature suggests that people prefer to be viewed as cooperative and trustworthy in social relationships (Schlenker & Weigold, 1990). This general self-presentation tendency is more salient in situations where individuals feel the need to conform to someone who holds a higher status (Schlenker & Leary, 1982). Given this, we expect victims of leader abuse with a relational self-concept to be motivated to preserve their self-image in the eyes of the leader. We thus hypothesize:
Hypothesis 3: Abusive supervision and relational self-concept will interact to predict image preservation motives, such that abusive supervision will have a significant positive effect on image preservation motives only in the presence of a stronger (vs. weaker) relational self-concept.
Similarly, we expect that relational self-concept serves as a catalyst for victims of leader abuse to act on their image preservation motives in order to actively craft a more favorable reputation in the eyes of the leader. For example, when victims of leader abuse are concerned with how their leader views them, they may be motivated to either employ a more active coping strategy (Harvey, Stoner, Hochwarter, & Kacmar, 2007; Wee et al., 2017) or to make a proactive effort to craft a more favorable self-image (Troester & Van Quaquebeke, 2021; Yu & Duffy, 2021). We propose that for those who hold a stronger relational self-concept, their experience of leader abuse will motivate them to engage in ingratiation toward the abusive leader as a result of their heightened concerns for image preservation. Whereas helping behaviors are hypothesized to target coworkers (hypothesis 2), we predict that those experiencing abusive supervision will respond with conforming behaviors characteristic of ingratiation when targeting their boss rather than with prosocial motives associated with helping behavior due to the lower emotional toll that ingratiation requires. Of interest here is the other-enhancing nature of ingratiation, which involves the use of compliments or flattery toward the leader (Bolino, Kacmar, Turnley, & Gilstrap, 2008; Liden & Mitchell, 1988). These behaviors may present an opportunity for victims of abusive supervision to actively “close the gap” and create more favorable impressions from the leader (Leary & Kowalski, 1990). Different from helping behavior toward the leader that has been shown to enhance a victim's feelings of self-worth following leader abuse (Vogel & Mitchell, 2017), we contend that the use of ingratiation tactics indicates victims’ focus on restoring a favorable image in the eyes of their leader. In other words, ingratiation could be viewed as a more explicit or purposeful act that expresses a person's willingness to reconcile the dysfunctional dyadic relationship.
Employee ingratiation is considered an effective self-presentation tactic (Jones & Pittman, 1982; Leary & Kowalski, 1990), and is often directed toward someone who is in control of and capable of providing valuable resources (Liden & Mitchell, 1988). Ingratiation can help cultivate interpersonal liking and is often used as an initial self-presentation strategy in developing interpersonal relationships (Aguinis, Nesler, Hosoda, & Tedeschi, 1994). Research has shown that employees’ use of ingratiation tactics positively predicts their leader's liking and performance evaluations (Wayne & Ferris, 1990), improves the quality of social exchange relationships with their superiors (Koopman et al., 2015), and reduces psychological stress resulting from ostracization from coworkers (Wu, Yim, Kwan, & Zhang, 2012). Since those with a stronger (but not weaker) relational self-concept are typically committed to strengthening social relationships at work (Johnson, Chang, & Yang, 2010), we argue that when experiencing leader abuse these individuals are likely to develop stronger image preservation motives toward their leader and to engage in ingratiation behavior to show their willingness to repair the dysfunctional dyadic relationship. That is, ingratiation can be portrayed as a self-presentation tactic through which victims of leader abuse actively craft a more favorable self-image in the eyes of their leader. For those with a weaker relational self-concept, however, we expect that they will not engage in self-image maintenance—such as using the ingratiation tactic—because they are less concerned about the interpersonal implications of having a dysfunctional relationship with their abusive leader and therefore should be unlikely to commit self-regulatory effort to repairing an already broken social relationship. With the above said, we hypothesize:
Hypothesis 4: Abusive supervision will have a positive indirect effect on employees’ ingratiation toward the leader via image preservation motives only in the presence of a stronger (vs. weaker) relational self-concept.
Overview of Methods
We conducted two field studies to test our model. In Study 1 we tested the identity-consistent self-image maintenance mechanism toward coworkers. In Study 2, we replicated our findings from Study 1 and tested the complete model.
Study 1
Method
Setting and procedures
Employees and their direct supervisors working at a large convenience store chain in China were invited to participate in this study. These participants were attending a company-wide workshop about leadership skills development. They came from 474 different work groups across organizational units (stores, logistics stations, regional operations centers). To alleviate the burden on supervisors, we picked one employee from each work group whose first name appeared first, alphabetically, on the personnel list of the work group. The invited participants worked in a number of areas such as retail sales, field operations, inventory management, accounting, human resources, marketing, customer service, digital merchandising, facilities, and support. Management indicated to us that invited participants had regular interactions with their supervisor at work.
Survey data were collected at three time points. At Time 1, we invited 474 employees (i.e., one per work group) to complete a survey about abusive supervision and relational self-concept. In all, 442 responses (93.2%) were collected. At Time 2, which took place 3 weeks after the completion of the first survey, we asked those 442 employee respondents to complete a second survey about their relational RMC. A total of 376 responses (85.1%) were collected. At Time 3, which was about 2 weeks after the completion of the Time 2 survey, we invited the direct supervisor of those 376 employee respondents who completed both Time 1 and Time 2 surveys to provide ratings of their subordinate's helping behavior. Altogether, 253 supervisors (67.3%) completed the survey. Therefore, our final sample consised of 253 non-nested supervisor-subordinate dyads. Among these 253 employee respondents, 118 (46.6%) were female. The average age and organizational tenure were 34.3 and 7.0 years, respectively.
Measures
To ensure the accuracy of translation of all survey materials from English to Chinese, we followed standard translation and back-translation procedures (Brislin, 1970). Unless specified otherwise, respondents indicated the extent to which they agreed or disagreed with each statement on a 5-point Likert scale (1 = “strongly disagree” to 5 = “strongly agree”).
Abusive supervision was measured with the 5-item (α = .89) shortened scale validated in Mitchell and Ambrose (2007) based on Tepper's (2000) original scale. Specifically, Mitchell and Ambrose (2007) identified two unique components of abusive supervision: passive-aggressive and active-aggressive. Following Mitchell and Ambrose (2007), we used only the active-aggressive dimension to represent abusive supervision, as we expected that the more active component of leader mistreatment would be more salient in triggering a person’s relational RMC. One sample item was: “This supervisor told me my thoughts or feelings were stupid.”
Relational self-concept was measured with the 5-item (α = .93) relational identity scale (Johnson, Selenta, & Lord, 2006). One sample item was: “If a friend was having a personal problem, I would help him/her even if it meant sacrificing my time or money.”
Relational RMC was measured with four items (α = .92) adapted from Baer et al.'s (2015) original RMC scale. The items were revised to reflect the relational aspect of reputation maintenance concerns. In the instructions, we explained to participants that “being a relational person,” as described in the items we used, means that someone not only values relationships but also can develop and maintain harmonious relationships with others. Items were: “I’m concerned about maintaining my image as a relational person,” “I worry about protecting my reputation as a relational person,” “I feel the need to preserve the opinion others have of me being a relational person,” and “I’m preoccupied with keeping others’ views that I am relational intact.” We validated this adapted measure of relational RMC in a separate study (see Appendix 1).
Helping behavior was measured using the 7-item scale (α = .91) developed by van Dyne and LePine (1998). We changed the referent to tailor it for supervisor ratings. One sample item was: “This employee helped others in this work group with their work responsibilities.”
Control variables
We controlled for trait negative affect which was found to influence a person's reactions following experience of leader mistreatment (Aquino, Lewis, & Bradfield, 1999; Lian et al., 2012). We measured trait negative affect using a 5-item scale (α = .91); a short-form of PANAS developed by Thompson (2007) and used in a number of prior studies investigating negative workplace experiences or engagement in prosocial behaviors (e.g., Tangirala & Ramanujam, 2012). Specifically, we asked employee participants to think about themselves and how they normally felt (e.g., upset, hostile).
Analytical strategy
We first conducted a series of confirmatory factor analyses (CFA) to assess the measurement validity of our hypothesized model. We then compared this baseline model with several alternative models that were theoretically plausible. In the next step, we tested our proposed moderated mediation model using structural equation modeling (SEM) via Mplus Version 6.12 (Muthén & Muthén, 2010), which allowed us to simultaneously assess all the paths while accounting for measurement errors. Specifically, following recent suggestions (Cortina, Markell-Goldstein, Green, & Chang, 2021), we assessed our model using the full latent approach to test interactions. Following recommendations from Hayes (2015), we further assessed the significance of the moderated indirect effects and the index of moderated mediation in SEM.
Results
Tests of response bias
We evaluated potential response bias between employee respondents whose supervisors completed the survey and those whose supervisors did not complete the survey. The results showed non-significant difference in age (F(1, 440) = .070, p = .791), gender (F(1, 440) = 1.526, p = .217), organizational tenure (F(1, 440) = .808, p = .369), abusive supervision (F(1, 440) = .324, p = .569), relational self-concept (F(1, 440) = .089, p = .765), or relational RMC (F(1, 440) = .601, p = .439).
Tests of measurement models
Prior to testing our hypotheses, we conducted a series of CFAs using Mplus Version 6.12 to assess the measurement validity of our five-factor (i.e., four primary variables and trait negative affect) baseline measurement model. The CFA results showed that our baseline model produced good fit: χ2 = 473.90, df = 289, CFI = .96, TLI = .95, RMSEA = .05, SRMR = .05. We then compared this baseline model with several alternative models that were theoretically plausible. To assess these alternative models, we combined the items of the following variables, respectively, and loaded them onto one latent factor: (1) relational self-concept and relational RMC (χ2 = 1248.54, df = 293, CFI = .79, TLI = .76, RMSEA = .11, SRMR = .13), (2) relational self-concept and helping behavior (χ2 = 1470.58, df = 293, CFI = .74, TLI = .71, RMSEA = .13, SRMR = .15), (3) relational RMC and helping behavior (χ2 = 1133.29, df = 293, CFI = .81, TLI = .79, RMSEA = .11, SRMR = .09), and (4) trait negative affect and relational RMC (χ2 = 1240.45, df = 293, CFI = .79, TLI = .76, RMSEA = .11, SRMR = .13). We further assessed the fit of a single-factor model in which we loaded all the items onto one latent factor: χ2 = 3914.14, df = 299, CFI = .19, TLI = .12, RMSEA = .22, SRMR = .25. The chi-square difference test showed that our baseline model produced a significantly better fit than the best competing model (i.e., the third alternative model): Δχ2 (4) = 659.39, p < .01. We thus decided to retain our baseline model and proceed to hypothesis testing.
Tests of hypotheses
Table 1 presents the means, standard deviations, and inter-correlations of study variables. We tested our model using SEM and summarized the SEM results in Table 2. Given that running SEM via the full latent approach (when testing interactions) in Mplus did not produce model fit indices, we performed a separate SEM via the partial latent approach (i.e., treating the interaction term as an observed instead of a latent variable; Cortina et al., 2021) in Mplus, which showed adequate fit of our model: χ2 = 496.15, df = 315, CFI = .96, TLI = .96, RMSEA = .05, SRMR = .05. We report the SEM results below using the full latent approach. Our findings held using the partial latent approach in SEM.
Descriptive Statistics and Correlations of Study Variables (Study 1)
Note: N = 253. RMC = reputation maintenance concerns. Gender was coded as 0 = female, 1 = male. Employee respondents reported on age, gender, organizational tenure, trait negative affect, abusive supervision, and relational self-concept at Time 1. Employee respondents further provided ratings of relational RMC at Time 2. The supervisor respondents provided ratings of helping behavior at Time 3.
*p < .05, two-tailed tests.
**p < .01, two-tailed tests.
Summary of Results From Structural Equation Modeling (Study 1)
Note: N = 253. RMC = reputation maintenance concerns. Results without the control variable included in the multilevel path analysis are reported in the parentheses. The change in R2 value was calculated based on the R2 value obtained from the full estimated model as presented and the R2 value obtained from a model in which the interaction term was not included.
*p < .05.
**p < .01.
As shown in Table 2, abusive supervision (β = .18, p = .063) was not significantly associated with relational RMC, while relational self-concept (β = .16, p = .020) positively predicted relational RMC. Also, there was a significant interactive effect of abusive supervision and relational self-concept on relational RMC (β = .45, p < .01), even after controlling for the positive impact of employees’ trait negative affect on their relational RMC (β = .18, p = .017). The change in r-squared value from adding in the interactive effect over a null model without the interaction effect was .089, showing an additional 8.9% of variance in relational RMC explained by the interaction of abusive supervision and relational self-concept. Moreover, the log-likelihood ratio test, which compares the relative fit of the full estimated model with the interaction effect over the null model without the interaction effect, proved to be significant (p < .01) based on the log-likelihood difference value (D = 10.064) and the difference in free parameters (one). We then plotted this significant interactive effect in Figure 2. We further conducted a simple slope t test to determine the significance of the slopes. The results showed that abusive supervision had a significantly positive effect on relational RMC under the condition of a stronger (i.e., +1SD) relational self-concept (b = .635, t = 4.651, p < .01). This relationship turned non-significant under the condition of a weaker (i.e., −1SD) relational self-concept (b = −.267, t = −1.956, p = .052). Hypothesis 1 was thus supported.
The Interactive Effect of Abusive Supervision and Relational Self-Concept on Relational Reputation Maintenance Concerns (Study 1)
Next, we evaluated the hypothesized moderated indirect effect proposed in Hypothesis 2. As shown in Table 2, relational RMC was positively associated with helping behavior (β = .38, p < .01). We further assessed the moderated indirect effect of abusive supervision on helping behavior in SEM and reported results of moderated indirect effects in Table 3. Specifically, the SEM results showed that abusive supervision had a significantly positive indirect effect on helping behavior (ρ = .229, 95% CI [.095, .363]) via relational RMC in the presence of a stronger (i.e., +1SD) relational self-concept. However, this indirect effect turned non-significant (ρ = −.089, 95% CI [−.189, .010]) in the presence of a weaker (i.e., −1SD) relationship identity. The difference between these two indirect effects was also significant (Δρ = .318, p < .01, 95% CI [.143, .494]). Following Hayes’s (2015) approach, we further assessed the index of moderated mediation (ω = .171, p < .01, 95% CI [.077, .266]) with helping behavior as the outcome variable. Hypothesis 2 was thus also supported.
Summary of Moderated Mediation Results From Structural Equation Modeling (Study 1)
Note: N = 253. The moderated mediation effects were assessed using SEM with the 95% confidence intervals (CI) also estimated. The CI that includes zero indicates that the estimated effect is not significant. The index of moderated mediation was assessed following Hayes’s (2015) approach. Results without the control variable included in SEM were reported in the parentheses.
Brief Discussion
In Study 1 we found support for Hypotheses 1 and 2 through a three-wave field study. Nevertheless, we conducted an additional field study (Study 2) to test the full model and address three potential limitations of Study 1. First, in Study 2 we operationalized abusive supervision by capturing both the active-aggressive and passive-aggressive components (Mitchell & Ambrose, 2007). Second, given that RMC is a relatively new construct, in Study 2 we provided further empirical evidence showing the convergent and discriminant validity of relational RMC. Finally, in Study 2 we collected data from a different context with nested supervisor-subordinate dyads, thus allowing us to use multilevel modeling technique to provide non-conflated estimates of within-person level indirect effects (Preacher, Zyphur, & Zhang, 2010).
Study 2
Method
Setting and procedures
Employees and their direct supervisors working at a large automobile manufacturing company in China were invited to participate in this study. The company focuses on Research and Development (R&D), manufacturing, and the distribution of electric and hybrid commercial vehicles and trucks. With the support of the Company's operations chief, we were given permission to invite employees working at a major manufacturing and R&D complex to participate in our survey study. These participants worked in a variety of areas such as mechanical solutions and integration, vehicle engineering, process design and engineering, quality engineering, production process engineering, embedded software systems engineering, administrative support, and operations. Our conversation with management suggested that these participants had regular interactions with their supervisor.
Survey data was collected at two time points. At Time 1, we invited 655 employees to complete a survey about abusive supervision, relational self-concept, relational RMC, image preservation motives, and trait negative affect. We collected 382 responses (58.3%). At Time 2, which was about 4 weeks after the completion of the Time 1 survey, we invited the direct supervisor of those 382 employee respondents to provide ratings of their subordinate's helping and ingratiation behaviors. Of the 94 supervisors we contacted, 67 returned a total of 272 responses, yielding a response rate of 71.3%. Therefore, the final sample consisted of 272 supervisor-subordinate dyads. Among those 272 employee respondents, 110 (40.4%) were female. The average age and organizational tenure were 37.1 and 8.4 years, respectively.
Measures
We again followed standard translation and back-translation procedures (Brislin, 1970). Unless specified otherwise, we asked respondents to indicate the extent to which they agreed or disagreed with the statements provided on a 7-point Likert scale (from 1 = “strongly disagree” to 7 = “strongly agree”). Relational self-concept (α = .96), relational RMC (α = .93), and helping behavior (α = .95) were all measured with the same scale used in Study 1. We also controlled for trait negative affect using the same 5-item scale (α = .92) from Study 1.
As we sought to account for the full-range implications of leader mistreatment, we measured abusive supervision with the complete 10-item scale (α = .87) validated by Mitchell and Ambrose (2007) based on Tepper's (2000) original scale. Sample items included: “This supervisor did not give me credit for jobs requiring a lot of effort” (passive-aggressive) and “This supervisor made negative comments about me to others” (active-aggressive).
Image preservation motives was measured with a 3-item scale (α = .89) used in Ng and Yam (2019) that captured a respondent's desire to preserve a favorable image of being creative at work. We adapted this scale by changing (a) the referent from others at work in general to the supervisor and (b) the time frame from experience of the present day to reflections on experience in the past month. Items included: “I felt I should display only job behaviors that made good impressions on my supervisor,” “I felt I should avoid behaviors that hurt my image at work in the eye of my supervisor,” and “I felt it was important for me to make a good impression on my supervisor at work.”
Ingratiation was measured by adapting the 8-item ingratiation tactic scale (α = .94) developed by Lee et al. (1999). We changed the referent of the original scale to make it suitable for supervisor ratings on their subordinates. Among the eight items, one item captured self-enhancing ingratiation while the other seven items captured other-enhancing ingratiation. Given the focus of our study on subordinates’ use of ingratiation tactics toward the supervisor (i.e., other-enhancing), we decided to drop the lone item that captured self-enhancing ingratiation (“This subordinate told me about his/her positive qualities”). The CFA results gave empirical support to our decision, as this self-enhancing ingratiation item failed to load onto the latent factor (with a factor loading of −.23). Sample items included: “This subordinate used flattery to win my favor,” “This subordinate expressed his/her opinions in a way that I would like,” and “This subordinate complimented me to get me on his/her side.”
Analytical strategy
We employed the same strategy used in Study 1 to assess the measurement validity of our proposed model. In the next step, given the nested nature of our data in Study 2, we chose to test our model using multilevel path analysis via Mplus Version 6.12 (Muthén & Muthén, 2010) following the general guidance from Preacher and colleagues (Preacher, Zhang, & Zyphur, 2011; Preacher et al., 2010). To justify the need to partition out within-person and between-person level effects in our analyses, we first calculated the ICC values (Bliese, 2000) of all the dependent variables: relational RMC (ICC1 = .20, ICC2 = .50), image preservation motives (ICC1 = .29, ICC2 = .63), helping behavior (ICC1 = .21, ICC2 = .52), and ingratiation (ICC1 = .29, ICC2 = .63). These ICC values showed notable variance at the between-person level in these four dependent variables, thereby supporting our decision to use multilevel modeling to analyze our data. This multilevel analytic approach helped reduce the potential confounding effects of contextual factors (i.e., between-person level effects) while accounting for the conflation of the within-person and the between-person components of the effects estimated (Preacher et al., 2011). Specifically, we followed recommendations from Preacher et al. (2011) and assessed our proposed relationships at both within-person and between-person levels (i.e., relying on the principles of the recommended MSEM analytical framework to conduct our multilevel path analyses), instead of simply accounting for the between-person effects of the predictor variables and focusing exclusively on the indirect effects assessed at the between-person level (Zhang, Zyphur, & Preacher, 2009). We further calculated the index of moderated mediation in Mplus following Hayes’s (2015) recommendations.
Results
Tests of response bias
Similarly, we evaluated potential response bias between employee respondents whose supervisors completed the survey and those whose supervisors did not complete the survey. The results showed non-significant difference in age (F(1, 380) = 2.112, p = .147), gender (F(1, 380) = .544, p = .461), organizational tenure (F(1, 380) = 1.298, p = .255), abusive supervision (F(1, 380) = 1.211, p = .272), relational self-concept (F(1, 380) = 1.963, p = .162), relational RMC (F(1, 380) = 2.498, p = .115), or image preservation motives (F(1, 380) = 1.074, p = .301).
Tests of measurement models
Given that our data was nested, we first followed recommendations from Dyer, Hanges, and Hall (2005) and Muthén (1994) to assess the covariance structure of our variables at the between-person level, and then attempted to conduct multilevel CFA to evaluate model fit at both within-person and between-person levels. However, given the relatively small size of clusters and the not-positive definite covariance matrix at the between-person level, our multilevel CFA on a seven-factor baseline measurement model (i.e., six study variables and trait negative affect) did not converge. Following suggestions from Ryu and West (2009), we attempted to assess as much information as possible about the psychometric properties at the between-person level in our multilevel CFA with model modifications specified based on the error messages received in Mplus. Eventually, a multilevel measurement model successfully converged after we (a) fixed both the item residual variance and factor variance to zero for image preservation motives, helping behavior, and ingratiation at the between-person level and (b) fixed the covariance to zero among all the variables at the between-person level. No modification was made at the within-person level. Results from multilevel CFA showed that this modified measurement model produced good fit: χ2 = 2223.62, df = 1549, CFI = .96, TLI = .96, RMSEA = .04, SRMR(within) = .05.
Although this modified multilevel measurement model allowed us to assess some parameters at the between-personal level, it was substantially constrained in its capacity to provide a comprehensive evaluation of the overall model fit at the between-person level. Consistent with suggestions from Ryu and West (2009), in the next step we focused our attention on the assessment of model fit at the within-person level. We first conducted a series of CFAs to assess the measurement validity of our seven-factor baseline measurement model. The CFA results showed that our baseline model produced good fit: χ2 = 1068.36, df = 758, CFI = .97, TLI = .97, RMSEA = .04, SRMR = .04. We then compared this baseline model with several alternative models that were theoretically plausible. To assess these alternative models, we combined the items of the following variables, respectively, and loaded them onto one latent factor: (1) relational RMC and image preservation motives (χ2 = 1667.86, df = 764, CFI = .91, TLI = .91, RMSEA = .07, SRMR = .08), (2) relational self-concept and relational RMC (χ2 = 2027.82, df = 764, CFI = .88, TLI = .87, RMSEA = .08, SRMR = .09), (3) relational self-concept and image preservation motives (χ2 = 1640.19, df = 764, CFI = .92, TLI = .91, RMSEA = .07, SRMR = .08), (4) relational RMC and helping behavior (χ2 = 1998.34, df = 764, CFI = .88, TLI = .87, RMSEA = .08, SRMR = .08), (5) image preservation motives and ingratiation (χ2 = 1577.08, df = 764, CFI = .92, TLI = .92, RMSEA = .06, SRMR = .07), and (6) trait negative affect, relational RMC, and image preservation motives (χ2 = 2645.48, df = 769, CFI = .82, TLI = .81, RMSEA = .10, SRMR = .12). We further assessed the fit of a single-factor model in which we loaded all the items onto one latent factor: χ2 = 8260.61, df = 779, CFI = .29, TLI = .25, RMSEA = .19, SRMR = .22. The chi-square difference test showed that our baseline model produced a significantly better fit than the best competing model (i.e., the fifth alternative model): Δχ2 (6) = 508.72, p < .01. We thus decided to retain our proposed baseline model.
Tests of hypotheses
Table 4 presents the means, standard deviations, and inter-correlations of the study variables. We tested the complete moderated mediation model via multilevel path analysis following the general guidance from Preacher et al. (2010, 2011), which addressed the problem of the unconflated multilevel modeling approach (Zhang et al., 2009) particularly in cases of moderated mediation testing, regarding the biased between-person effects toward the corresponding within-person effects. In our analyses, we also controlled for the potential confounding effect of trait negative affect on the mediators and the outcome variables at both within-person and between-person levels. The results from multilevel path analysis showed that our model produced good fit: χ2 = 17.18, df = 12, CFI = .98, TLI = .93, RMSEA = .04, SRMR (within) = .03. We summarized the results of path coefficients from multilevel path analysis in Table 5. To demonstrate the robustness of our findings, we also reported results without the control variable included in Table 5.
Descriptive Statistics and Correlations of Study Variables (Study 2)
Note: N = 272. RMC = reputation maintenance concerns. Gender was coded as 0 = female, 1 = male. Employee respondents reported on age, gender, organizational tenure, trait negative affect, abusive supervision, relational self-concept, relational RMC, and image preservation motives at Time 1. The supervisor respondents provided ratings of helping behavior and ingratiation at Time 2.
*p < .05, two-tailed tests.
**p < .01, two-tailed tests.
Summary of Path Coefficients From Multilevel Path Analysis (Study 2)
Note: N = 272. RMC = reputation maintenance concerns. Results without the control variable included in the multilevel path analysis were reported in the parentheses. The change in R2 value was calculated based on the R2 value obtained from the full estimated model as presented, and the R2 value obtained from a model in which the interaction term was not included.
*p < .05.
**p < .01.
We first sought to replicate findings from Study 1 regarding tests for Hypotheses 1 and 2. As shown in Table 5, both abusive supervision (γ = .15, p < .01) and relational self-concept (γ = .17, p < .01) were positively associated with relational RMC. As expected, abusive supervision and relational self-concept interacted to predict relational RMC (γ = .20, p < .01). The change in r-squared value resulted from adding in the interactive effect over a null model without the interaction effect was .086, showing an additional 8.6% of variance in relational RMC explained by the interaction of abusive supervision and relational self-concept. We plotted this significant interactive effect in Figure 3. Results from the simple slope t test showed that abusive supervision was positively associated with relational RMC (b = .350, t = 5.633, p < .01) in the presence of a stronger (i.e., +1SD) relational self-concept. This relationship turned non-significant (b = −.050, t = −.805, p = .422) under the condition of weaker (i.e., −1SD) relational self-concept. Therefore, we also found support for Hypothesis 1 in Study 2.
The Interactive Effect of Abusive Supervision and Relational Self-Concept on Relational Reputation Maintenance Concerns (Study 2)
We further reported results of moderated indirect effects in Table 6. Specifically, results from multilevel path analysis showed that abusive supervision had a significant and positive indirect effect on employees’ helping behavior (ρ = .134, p < .01, 95% CI [.064, .204]) via relational RMC in the presence of a stronger (i.e., +1SD) relational self-concept. This indirect effect turned non-significant (ρ = −.043, p = .060, 95% CI [−.089, .002]) in the presence of a weaker (i.e., −1SD) relational self-concept. The difference between these two indirect effects was also significant (Δρ = .178, p < .01, 95% CI [.084, .271]). Following Hayes’s (2015) approach, we further assessed the index of moderated mediation (ω1 = .060, p < .01, 95% CI [.028, .092]) with helping behavior as the outcome variable via multilevel path analysis. These results provided support for Hypothesis 2 in Study 2.
Summary of Moderated Mediation Results From Multilevel Path Analysis (Study 2)
Note: N = 272. RMC = reputation maintenance concerns. CI = confidence interval. The moderated mediation effects were assessed in multilevel path analysis following recommendations from Preacher et al. (2010, 2011). The index of moderated mediation was assessed following Hayes’s (2015) approach. The CI that includes zero indicates a non-significant estimated effect. Results without the control variable included in the multilevel path analysis were reported in the parentheses.
Hypothesis 3 proposed that abusive supervision and relational self-concept would interact to predict image preservation motives. As reported in Table 5, abusive supervision had a significant main effect on image preservation motives (γ = .29, p < .01), while relational self-concept also positively predicted image preservation motives (γ = .17, p < .01). As expected, abusive supervision and relational self-concept interacted to predict image preservation motives (γ = .26, p < .01). The change in r-squared value from adding in the interactive effect over a null model without the interaction effect was .208, showing an additional 20.8% of variance in relational RMC explained by the interaction of abusive supervision and relational self-concept. We plotted this significant interactive effect in Figure 4. Results from the simple slope t test showed that abusive supervision was positively associated with image preservation motives (b = .550, t = 10.437, p < .01) in the presence of a stronger (i.e., +1SD) relational self-concept. This relationship turned non-significant (b = .030, t = .569, p = .570) under the condition of a weaker (i.e., −1SD) relational self-concept. Hypothesis 3 was thus supported.
The Interactive Effect of Abusive Supervision and Relational Self-Concept on Image Preservation Motives (Study 2)
Similar to the procedures we took in Study 1, we also performed the log-likelihood ratio test comparing the relative model fit of our main model with the two interactive effects over the null model without the two interactive effects. The log-likelihood ratio test was significant (p < .001) based on the log-likelihood difference value (D = 51.979) and the difference in free parameters (two). This yielded further support for the significant portion of additional variance explained by the inclusion of the two interaction effects (i.e., abusive supervision interacting with relational self-concept to predict both relational RMC and image preservation motives).
Hypothesis 4 proposed that there would be a positive and stronger indirect effect of abusive supervision on employees’ ingratiation toward the leader via image preservation motives in the presence of a stronger (vs. weaker) relational self-concept. As reported in Table 6, abusive supervision had a positive indirect effect on employees’ ingratiation (ρ = .207, p < .01, 95% CI [.120, .294]) via image preservation motives in the presence of a stronger (i.e., +1SD) relational self-concept. This indirect effect turned non-significant (ρ = −.027, p = .165, 95% CI [−.065, .011]) in the presence of a weaker (i.e., −1SD) relational self-concept. The difference between these two indirect effects was also significant (Δρ = .234, p < .01, 95% CI [.131, .336]). Similarly, following Hayes’ (2015) approach, we assessed the index of moderated mediation (ω1 = .079, p < .01, 95% CI [.044, .114]) with ingratiation as the outcome variable via multilevel path analysis. Therefore, we found support for Hypothesis 4.
General Discussion
Theoretical Implications
Our research makes at least three important contributions to the extant literature. First, our study contributes to our understanding of the self-appraisal process following leader abuse by breaking new theoretical ground and framing this process as a person's identity-driven self-image maintenance that unfolds via self-presentation tactics. Recent studies have explained victims’ heightened concern for self-image as reflected in their emotional appraisal (e.g., self-blame) of leader mistreatment (Troester & Van Quaquebeke, 2021; Yu & Duffy, 2021). We contend that, although a person's self-image maintenance following leader abuse may trigger episodic emotional reactions (Oh & Farh, 2017), this self-concept driven process is more than just emotional in nature. Indeed, coping with an abusive boss may involve more comprehensive evaluation of coping strategies, a process that is often thoughtfully planned and executed. For example, Wee et al. (2017) reported that employees relied on approach balancing strategies, such as coalition formation or value enhancement, to help create leader perceptions of dependence on them, therefore attenuating subsequent leader mistreatment. We address this need for a more comprehensive understanding about the psychological mechanisms involved as one copes with an abusive boss, and present a novel theoretical perspective explaining such mechanisms.
Our investigation of self-image maintenance toward both coworkers and the leader is needed because recent studies have found that individuals might address their self-image concerns following leader abuse (e.g., self-blaming emotions) by engaging in either OCB toward peers (Yu & Duffy, 2021) or helping behavior toward the leader (Troester & Van Quaquebeke, 2021). We enrich these emerging theoretical perspectives, which almost exclusively focus on the social regulation of emotions or emotional appraisals, by providing a self-presentation and self-concept orientation view in our study. As such, we provide a more complete look at a person’s identity-driven self-image maintenance process with the self-presentational outcomes toward both coworkers and the leader. This is much needed as recent research suggests that victims’ episodic emotional reactions following leader abuse are often coupled with more thoughtful and complicated evaluations of self-image driven by more stable deeper-level factors (Oh & Farh, 2017). Our study therefore provides a novel theoretical alternative to account for victims’ self-image reevaluation. Our findings show that for those with a stronger relational self-concept, relational RMC and image preservation motives functioned as important mediating mechanisms representing such heightened concerns—and therefore stronger motivations—for preserving a favorable and socially desirable self-image in the eyes of coworkers and the abusive leader. This active self-image maintenance is necessary because the dysfunctional nature of the abusive leader-subordinate interactions not only signals relational exclusion from the abusive leader, but also communicates to the broader social context the victims’ inability to maintain relationships (Priesemuth et al., 2014; Schaubroeck et al., 2016).
We built on prior research on RMC (Baer et al., 2015) to introduce the concept of relational RMC to study the broader social impact of leader abuse because victims’ concern associated with anticipated loss of reputation is largely overlooked in the literature. We help inform abusive supervision research regarding the need and the potential to move beyond the social justice, emotional appraisal, and self-regulation perspectives more commonly used in the study of victims’ perceptions of third-party reactions to leader abuse (Mitchell et al., 2015; Shao, Li, & Mawritz, 2018; Xu et al., 2020). The role that relational self-concept plays in our model demonstrates the importance of identity beliefs in predicting victims’ reactions to leader abuse, which has been absent in prior examinations. In addition, by focusing on image preservation motives toward the leader as an important mediator, we supplement prior research which has positioned victims’ use of self-presentation tactics toward the abusive boss as a result of diminished self-esteem (Vogel & Mitchell, 2017).
Second, recent studies (e,g, Troester & Van Quaquebeke, 2021; Yu & Duffy, 2021) have called for attention to individuals’ trait-like characteristics that could account for variation in desires for self-image protection following leader abuse. We answer these calls by showing that identity beliefs may be a key factor driving self-image maintenance following leader abuse. Our findings show that in the context of abusive supervision those with a stronger relational self-concept (Brewer & Gardner, 1996) were more likely to sense the need for self-image maintenance (Tepper, 2000; Tepper et al., 2017) and act to create a more favorable and socially desirable self-image. Our study extends prior research which established that those with a relational self-concept are more sensitive to hostile interpersonal encounters because they challenge their deeply held identity beliefs (Schilpzand & Huang, 2018) by showing that these individuals take socially beneficial action addressing the need for self-image protection. Our findings also add to growing conversations about the application of self-presentation perspectives in self-image protection (e.g., Bonner, Greenbaum, & Quade, 2017; Wakeman, Moore, & Gino, 2019).
Finally, while recent research relying on the emotional appraisal perspective offers encouraging findings about socially beneficial reactions from victims of leader abuse toward either the observer (i.e., coworkers) or the source (i.e., leader) of abuse (Troester & Van Quaquebeke, 2021; Yu & Duffy, 2021), our study adds to this line of scholarly conversation by showing the dual mediating mechanisms through which the abused employees seek to maintain their identity-consistent self-image toward both the observer and the source of the abuse simultaneously. Our research thus adds nuanced understanding about the potentially relationship-promotive consequences of leader abuse to the abusive supervision literature, which has largely shown that a person's experience of abusive supervision negatively influences their interpersonal behaviors, such as deviance toward the leader (e.g., Lian et al., 2014a; Mitchell & Ambrose, 2007) or coworkers (Lian et al., 2012). Importantly, our findings provide empirical support to the notion that victims’ episodic emotional coping following leader abuse does not necessarily lead to negative outcomes and may surprisingly create conditions for the parties involved to find a way to “break the cycle” of leader mistreatment (Oh & Farh, 2017; Wee et al., 2017). As a result, answering the call from Wee et al. (2017), our research takes a step forward in uncovering victims’ conscious, deliberate, or even calculative efforts in their attempt to repair the dysfunctional dyadic relationship with their abusive leader.
Practical Implications
Our study shows how the victims of abusive supervision may “manage around” to mitigate the damaging effects of leader mistreatment on their relational reputation in the work group and “manage up” to solicit more favorable views from the abusive boss. These seemingly favorable outcomes of abusive supervision, however, come at a significant cost to the follower from a psychological perspective, especially to those who are concerned with maintaining their relational reputation. For these individuals, the experience of abuse at work triggers compensatory behaviors that seek to counteract the potential damage that their involvement in an abusive relationship may cause to their relational reputation.
Victims of abuse often internalize their suffering and may even tend to blame themselves for what they are experiencing (Troester & Van Quaquebeke, 2021; Yu & Duffy, 2021). In this paper we reveal a reason why victims of workplace abuse not only continue in the abusive relationship but also may engage in positive behaviors (ingratiation and helping) as a result. Unfortunately, the follower's desire to shore up a threatened relational identity through these positive behaviors may actually end up reinforcing the leader's abusive behavior because they see positive outcomes from the abuse. Therefore, while we absolutely acknowledge the many and varied reasons why some individuals tolerate abusive work relationships (job insecurity, financial strain, lack of other options, etc.), we encourage those who are experiencing abuse in the workplace to consider taking actions to stop the abuse if at all possible, even if it means that their workplace reputation may suffer. Actions that can be taken to stop abuse at work include speaking with your leader's boss, filing a formal complaint with human resources, asking to be reassigned to another leader, or securing employment elsewhere. Especially for those who are concerned with their relational reputation at work, calling more attention to the abuse they are experiencing may be a difficult step, but it may be the only way to stop the cycle of abuse.
Limitations and Future Research Directions
Our study, while possessing a number of strengths, is not without limitations. First, it is possible that the effects we report can be influenced by how these abused employees are treated by others in the work unit. In particular, prior research has shown that employees suffering from uncivil coworker treatment tend to feel socially isolated (Schilpzand & Huang, 2018) and care less about the workplace environment (Lim, Cortina, & Magley, 2008). These unpleasant personal encounters with coworkers could demotivate abused employees from engaging in prosocial behaviors. Future research could explore how abused employees’ relationships with coworkers interact with relational self-concept to influence helping behavior.
Second, because it is beyond the scope of the present study, our model does not address how leaders may react to the impression management tactics from abused employees. Prior research found that leaders’ own self-regulation impairment could motivate them to enact interpersonal mistreatment (Mawritz, Greenbaum, Butts, & Graham, 2017). This could be an interesting direction for future research as leaders’ self-regulation impairment might significantly influence the effectiveness of abused employees’ ingratiation tactics, as those leaders with a strong self-regulatory capacity might be more likely to acknowledge such self-image maintenance attempts from abused employees and thus choose to repair the dysfunctional leader-follower relationship. In addition, future research might also benefit from an examination of leaders’ individual characteristics that could influence their response to victims’ attempts to repair the dyadic relationship. For example, managers who are more sensitive to signals of self-presentation tactics might be more likely to recognize such goodwill in their interpersonal interactions with the abused employees and thus respond more favorably.
Third, although we provide a novel theoretical perspective explaining employees’ prosocial responses to abusive supervision, we are limited by our research design in making definitive statements about the explanatory power of our studied mediating mechanisms over other established mediating mechanisms (e.g., self-esteem; Vogel & Mitchell, 2017 or emotional feelings of guilt; Yu & Duffy, 2021). That said, we did control for trait negative affect to try to account for potential trait-level predictors of victims’ emotional reactions following leader abuse. Future research may consider testing these mediating mechanisms simultaneously to examine their predictive power.
Finally, we collected our field data from China, a society where cultural factors could potentially lead to different interpretations of leader behaviors (Farh, Earley, & Lin, 1997). For example, previous research has suggested that individuals who hold traditional values are more likely to perceive experience of leader mistreatment as the leader executing authority (Liu, Kwan, Wu, & Wu, 2010). However, research findings have also shown that the new generation in China, which has witnessed significant industrialization at the societal level, tends to hold values more consistent with the Western culture than with traditional Chinese values (Ralston, Egri, Stewart, Terpstra, & Kaicheng, 1999). That said, the fact that our findings have shown significant moderated indirect effects is even more encouraging as such effects may be even more evident in a Western cultural setting where individuals feel more uncomfortable about being mistreated by their leader. We call for future research to explore this phenomenon in the Western context while also considering other societal and cultural factors that might influence how employees interpret their experience of abusive supervision.
Conclusion
We join emerging research on abusive supervision that investigates employees’ self-appraisal processes following leader abuse by theorizing, and testing through two field studies, an identity-driven self-image maintenance framework that explains why and under what conditions employees’ experience of leader abuse triggers concerns for preserving a favorable self-image and results in socially beneficial behaviors targeting coworkers and the leader. Our theoretical framework, which integrates self-presentation and self-concept orientation perspectives, supplements research that has relied on emotional appraisal perspectives or the social functional regulation of emotions. Our findings also call for more empirical examinations to uncover how leaders and subordinates may both contribute to break the cycle of leader abuse and instead foster a healthier and more productive dyadic interpersonal relationship.
Footnotes
Appendix 1. Construct,Convergent,and Discriminant Validity of Relational RMC
To validate the construct of relational RMC, we collected online survey data via Prolific Academic to empirically test for its construct, convergent, and discriminant validity. We published our survey study on the Prolific platform and set the expected sample size to 500. We also specified the following pre-screening eligibility criteria in Prolific: (1) 18 years and older; (2) currently work on a full-time job in the United States; and (3) work in the current organization for at least 6 months. We paid each participant $1.4 USD for completing a brief online survey. We received a total of 468 responses. Among those 468 responses, 11 were invalid and therefore excluded from our analyses due to failed response check. Therefore, our final sample consisted of 457 respondents, yielding a valid response rate of 91.4%. Among the 457 respondents, 206 (45.1%) were female. The average age and organizational tenure were 35.7 and 6.9 years, respectively. Some 324 participants (70.9%) reported having at least a college degree, and 426 (93.2%) self-identified as White Caucasian.
We performed confirmatory factor analysis (CFA) to evaluate the construct validity of relational RMC. We then compared relational RMC with each of the three relevant constructs: (a) RMC (the original version developed in Baer et al., 2015); (b) relationship-induced stress; and (c) feelings of social isolation. We argue that relationship-induced stress and feeling of social isolation are conceptually relevant constructs as they both capture psychological pressure from interpersonal interactions (Cacioppo & Cacioppo, 2014; Penney & Spector, 2005). Specifically, RMC (α = .89) was measured with the original four items used in Baer et al. (2015). Relationship-induced stress (α = .85) was measured by adapting four items from Motowidlo, Packard, and Manning (1986). We changed the referent of the four items to make it clear that the source of stress was about dealing with interpersonal relationships. We asked participants to reflect upon their social interactions at work with peers in the past month and rate on the following four items: “I feel a great deal of stress because of the interpersonal relationships I have to manage at work,” “Very few stressful things regarding managing interpersonal relationships happen to me at work” (reverse coded), “Managing interpersonal relationships at work is extremely stressful,” and “I almost never feel stressed about managing interpersonal relationships at work” (reverse coded). Feeling of social isolation (α = .94) was measured using the seven-item “intimate others” subscale from the UCLA-R Workplace Loneliness Scale (Russell et al., 1980). We asked participants to reflect only on their workplace interactions with peers in the past month and rate the following seven items: “I feel isolated from others,” “I feel left out,” “People are around me but not with me,” “I lack companionship,” “No one really knows me well,” “There is no one I can turn to,” and “I am no longer close to anyone.”
Regarding tests of construct validity, we performed CFA on the construct of relational RMC (α = .94). The CFA results showed that our construct of relational RMC produced acceptable fit: χ2 = 4.41, df = 2, CFI = 1.00, TLI = 1.00, RMSEA = .05, SRMR = .01. Therefore, we found support for the construct validity of relational RMC.
Regarding tests of convergent validity, the correlations results showed that relational RMC were positively related to all these three variables: RMC (r = .53, p < .01), relationship-induced stress (r = .20, p < .01), and feeling of social isolation (r = .17, p < .01), thus providing support for convergent validity of relational RMC with all these three constructs.
Finally, we also conducted a series of CFAs to assess if relational RMC would discriminate against all these three conceptually relevant constructs. The results showed that the measurement model in which we loaded items of relational RMC and RMC onto two unique latent factors produced a good fit: χ2 = 49.72, df = 19, CFI = .99, TLI = .98, RMSEA = .06, SRMR = .02. We then loaded items of relational RMC and RMC onto one single latent factor and obtained the following results: χ2 = 716.97, df = 20, CFI = .76, TLI = .66, RMSEA = .28, SRMR = .15. Results from the chi-square difference test showed that the former measurement model (i.e., the one in which relational RMC and RMC were treated as two unique latent factors) produced a significantly better fit: Δχ2 (1) = 667.25, p < .01. Similarly, the measurement model in which we loaded items of relational RMC and relationship-induced stress onto two unique latent factors (χ2 = 95.17, df = 19, CFI = .97, TLI = .96, RMSEA = .09, SRMR = .04) produced a significantly better fit (Δχ2 (1) = 848.52, p < .01) than the one in which we loaded items of these two constructs onto one single latent factor (χ2 = 943.69, df = 20, CFI = .65, TLI = .51, RMSEA = .32, SRMR = .21). Finally, the measurement model in which we loaded items of relational RMC and feeling of social isolation onto two unique latent factors (χ2 = 124.92, df = 43, CFI = .98, TLI = .98, RMSEA = .07, SRMR = .03) produced a significantly better fit (Δχ2 (1) = 2337.82, p < .01) than the one in which we loaded items of these two constructs onto one single latent factor (χ2 = 2462.74, df = 44, CFI = .42, TLI = .27, RMSEA = .35, SRMR = .34). These results therefore provided full support for the discriminant validity of relational RMC against three conceptually relevant yet psychometrically distinct constructs (i.e., RMC, relationship-induced stress, and feeling of social isolation).
Acknowledgements
We are indebted to Associate Editor, Rebecca Greenbaum, and the two anonymous reviewers for their very helpful comments and guidance throughout the review process. We would like to acknowledge Huiwen Lian and Jeremy Mackey for their feedback on earlier versions of this manuscript. We are also grateful for Zhen Zhang, Dina Krasikova, and Eunsook Kim for their insights on methodologies.
