Ripple effects of supervisor counterproductive work behavior directed at the organization: using affective events theory to predict subordinates’ decisions to enact CWB

Exposure to others' counterproductive work behaviors may significantly impact employees at work. However, research has yet to thoroughly examine third-party reactions to CWB, particularly supervisor CWB. We build on affective events theory, research regarding supervisors' role in shaping work experiences, and research on vicarious effects of CWB to better understand how exposure to supervisor CWB-O can influence subordinates. Based on insights from preliminary studies, the within-person vignette-based focal study (N = 1232 ratings nested in 176 raters) assessed when and how supervisor CWB-O predicts subordinates' decisions to enact their own CWB. Results suggest that supervisor CWB-O enacted in the context of a negative supervisor-subordinate relationship as well as severe, frequent supervisor CWB-O elicits negative subordinate reactions leading to greater intentions to enact CWB themselves.

Counterproductive work behaviors (CWB) refer to intentional acts on the part of employees that run counter to the legitimate interests of the organization or those within the organization (Sackett, [52]). Historically, the CWB literature has been dominated by two major research paradigms. Research taking the actor perspective explores factors that motivate individual enactment of CWB such as negative emotions (Spector & Fox, [60]), opportunity (Shoss, Jundt, Reynolds, & Kobler, [56]), and even revenge (Skarlicki & Folger, [57]). In contrast, research taking the target perspective captures the well-being and performance consequences that occur for those targeted or victimized by these behaviors (Hershcovis, [23]). Recently, however, researchers have expanded the CWB research domain to include an observer perspective that examines how those who are not primarily targeted but are exposed to these potentially harmful behaviors respond (Ferguson & Barry, [16]; Robinson, Wang, & Kiewitz, [48]). This work has tended to focus on reactions to coworker's misbehavior (Ferguson & Barry, [16]; Porath & Erez, [44]; Robinson et al., [48]) as well as vicarious effects of the ways that supervisors interpersonally treat other employees (e.g., abusive supervision; Harris, Harvey, Harris, & Cast, [22]; Mitchell, Vogel, & Folger, [40]).

In contrast, we know very little about how subordinates respond to how the supervisor behaves with regards to the activities of the organization, and particularly the consequences of organization-directed counterproductive work behaviors such as using work time for personal matters, sabotaging projects, or falsifying reports (Spector et al., [61]). This is surprising as Ferguson and Barry ([16]) point out that managers "are in a unique position to have a far-reaching or even ripple effect on the behavior of the organization's employees" (p. 90). Stories emerging from the Wells-Fargo scandal cited the misbehavior of managers as a source of bottom-line employees' fraud (Egan, [15]), and similarly, Enron employees pointed to the actions of management in driving wide-spread misbehavior of employees (Bartlett & Glinska, [5]). While research on interpersonal misbehaviors of supervisors (which are particularly personal as they are directed at fellow human beings) supports the idea that employees attend and respond to their supervisor's behavior, it tells us very little about how they might respond to supervisor's organization-directed misbehavior (which is less personal but still relevant to the employee's work experience), the features of the supervisor's behavior that may shape reactions, or whether or why these reactions may lead employees to enact their own harmful behavior. The nomological distinction in the literature between CWBs directed at individuals and CWBs directed at the organization (e.g., Robinson & Bennett, [47]; Spector et al., [61]) serves as the conceptual basis for examining the particular effects of supervisor CWB-O on employees.

Specifically, the current study examines the connection between features of a supervisor's CWB directed toward the organization (CWB-O) and the employee's intentions to subsequently engage in various acts of misbehavior themselves. Building from affective events theory (AET; Weiss & Cropanzano, [63]) as well as recent theory that emphasizes the emotionality attached to supervisor behaviors (Cropanzano, Dasborough, & Weiss, [10]), we purport that certain characteristics of supervisor CWB-O elicit negative cognitions and emotions, which will, in turn, contribute to subordinates' intentions to enact their own CWB. In particular, we distinguish three characteristics that may influence subordinates' reactions: frequency of the supervisor's behavior, severity of the supervisor's behavior, and quality of the relationship between subordinate and supervisor.

Our study advances theory and research pertaining to CWB in several important ways. First, it advances research on the impact of third-party exposure to CWB-O by demonstrating how certain features of supervisors' organization-directed CWBs impact subordinates' appraisals, emotions, and ultimately own intentions to enact CWB. While social-learning (Bandura & Walters, [4]) and social information processing (Salancik & Pfeffer, [53]) theories have provided valuable foundations to help explain why employees may mimic misbehaviors they see at work, the absence of emotion processes in these theories means we are missing a critical mechanism in understanding employee enactment of CWBs. Given that negative emotions play a primary role in CWB enactment decisions (Rodell & Judge, [49]; Spector & Fox, [60]), understanding how supervisor CWB-O plays a role in triggering an emotional process can give us more insight into another possible mechanism for the permeation of CWB in organizations from supervisors to subordinates. Thus, we leverage AET to examine the important roles of both features of the social situation and emotion in shaping whether and when subordinates enact CWB when their supervisor first does so. Importantly, we help to explain an intriguing paradox – that subordinates' negative views of their supervisor's CWB are associated with greater intentions to enact CWB themselves. Second, we link observer and actor perspectives of CWB-O by examining how and when subordinates' reactions to supervisor CWB-O impact their intentions to enact CWB themselves. Finally, we identify and examine specific event features surrounding supervisors' CWB-O that can further enhance the impact of these behaviors on subordinates in the workplace. In doing so, we establish an empirical foundation to understand whether and how CWB not targeted at particular individuals can still impact those in the workplace. These characteristics provide further insight on the potential ripple effect of supervisor CWB-O and suggest practical implications.

AET takes an event-based approach toward understanding behavior in the workplace. Specifically, AET posits that workplace events elicit emotions that drive affectively laden behavior (Weiss & Cropanzano, [63]). Particularly relevant for this process is the subordinate's appraisal of a workplace event. An event that is appraised as unjust, negative, or stressful is more likely to elicit an emotional reaction of anger, frustration, and/or anxiety (Dewe & Trenberth, [13]; Weiss, Suckow, & Cropanzano, [64]) than an event that is appraised as non-threatening or positive (Lazarus & Folkman, [35]; Weiss & Cropanzano, [63]). In turn, these affective experiences, including both cognitive appraisals and subsequent-associated emotions, shape emotion-driven behavior. In addition, AET holds that employees' affective dispositions contribute to their emotional reactions and affect driven behaviors. For example, individuals with trait negative affectivity or neuroticism may be predisposed to react more strongly to negative events; thus, it is important to include dispositions along with affective reactions to events (Weiss & Cropanzano, [63]). AET has recently been positioned as an explanatory model for how subordinates react to their leaders, suggesting that behaviors of the supervisor represent "events" to which subordinates then appraise and react (Cropanzano et al., [10]). AET has also been utilized to understand the genesis of CWB in organizations, where research suggests that negative emotion-eliciting events can lead to the enactment of CWB (Rodell & Judge, [49]).

We extend theory and research surrounding AET by suggesting that a supervisor's CWB-O serves as an emotion-eliciting event that can trigger subordinates' decisions to engage in CWB. Consistent with work that suggests that events can be meaningfully distilled and examined based on their features (e.g., Butts, Becker, & Boswell, [9]), we expect that features of supervisor CWB-O (e.g., severity of the CWB) shape subordinates' negative appraisals, which in turn elicit negative emotions and subsequent decisions to enact their own CWB. We focus specifically on negative cognitions and emotions as reactions to and drivers of CWB as the emotion-centered model of work behavior specifically theorizes the direct connection between negative emotions and CWB (Spector & Fox, [60]; see also Rodell & Judge, [49]). Figure 1 presents the full model.

Graph: Figure 1. Overall hypothesized model.

There are three different ways that subordinates are targeted with, influenced by, and otherwise experience CWB in the workplace: direct impact, vicarious impact, and ambient impact (Robinson et al., [48]). Direct impact refers to an employee being the direct target of deviant behavior. In contrast, vicarious and ambient impact focus primarily on the bystander and environmental elements of observed CWB, respectively. Vicarious impact refers to witnessing or learning of another's deviant behavior. In contrast, ambient impact indicates a working environment where deviant behavior is the norm (Robinson et al., [48]). Recognizing that ambient impact assumes the effects of deviant behavior as a macro-level norm rather than the explicit observations of an individual required of vicarious impact, we have specifically targeted vicarious events as our main focus in the current paper.

Specifically, we suggest that supervisor-enacted CWB-O serves as a vicarious event that may elicit negative appraisals and negative emotional and behavioral reactions. Because supervisors serve as representatives of the organization, subordinates are particularly attuned to their supervisor's behavior (Shoss, Eisenberger, Restubog, & Zagenczyk, [54]). Recent insights suggest this is the case even if the supervisor's behavior is not directly targeted at the subordinate (Robinson et al., [48]). Along these lines, research suggests that negative supervisor behaviors serve as significant emotion-eliciting events for subordinates (Cropanzano et al., [10]) and that subordinates may be more likely to recall negative experiences with their supervisors over positive ones (Dasborough, [11]). That is, negative supervisor behaviors are particularly salient events to subordinates.

In particular, we suggest that supervisor CWB-O serves as an emotion-eliciting event prompting subordinates' negative appraisals, and subsequent negative emotions and negative acts. Supervisors' CWB-O violates subordinates' expectations for how the supervisor should behave as a representative of the organization, as someone entrusted with higher power and status, and as someone who will evaluate the subordinate's performance. Indeed, prior research has found that when supervisors engage in inappropriate behaviors, subordinates react with a variety of negative emotions, including frustration, anger, and annoyance (Dasborough, [11]). For example, Rupprecht, Kueny, Shoss, and Metzger ([51]) found that negative emotions mediated the relationship between incongruence of expectations of supervisor behaviors and supervisors' actual behaviors, and subordinates' own CWB. Thus, we assess subordinates' appraisals of supervisors' behavior to determine the extent to which they believed the behavior to be negative, stressful, unjust, and incompetent.

To the extent that subordinates develop negative appraisals of a supervisor's CWB, we anticipate they will experience negative emotions and, via their negative emotions, develop intentions to enact their own CWB. Such predictions are consistent with AET. Further, theory and research on CWB suggest that CWB is the outcome of a process wherein negatively appraised events stimulate negative emotion, which in turn results in CWB (e.g., Rodell & Judge, [49]). Additionally, narrower emotion-focused theories (i.e., the emotion-centered model of work behavior; Spector & Fox, [60]) have been used consistently to link negative emotions to CWB (e.g., Fox, Spector, & Miles, [18]; Matta, Erol‐Korkmaz, Johnson, & Biçaksiz, [37]). From this perspective, CWB is an emotion-driven behavior enacted in response to employees' negative experiences at work.

Further, AET emphasizes that features of events shape individuals' appraisals, and in turn, their emotional responses, and that these features are critical to understanding how events drive reactions (Weiss & Cropanzano, [63]). Thus, we took both an inductive and deductive approach to identifying event features of supervisor CWB-O based on the extant literature on third-party reactions to CWB and based on an initial interview study to generate information about the phenomena (see preliminary study 1 discussed below). The features that appeared in the literature as well as in our interview study included frequency of the CWB-O, severity of the CWB-O, and relationship quality between subordinate and supervisor.

We anticipate that more frequent supervisor-enacted CWB-O will lead to negative appraisals and reactions on the part of employees. When individuals recall an emotional experience, most do not simply report a single event followed by a single emotion. Instead, as AET posits, individuals usually report a series of emotional transactions with their environment with an underlying theme (Frijda, [19]; Weiss & Cropanzano, [63]). Thus, frequently witnessing a supervisor engage in CWB (e.g., everyday) may lead employees to have an increasingly negative environmental perception. Furthermore, according to excitation transfer theory (Zillmann, [65]), emotional experiences are enhanced when individuals are in aroused states. For instance, if an individual is already in a poor mood following one event, then subsequent negative events may cause an acceleration in their negative emotions. Although little research has focused on CWB-O, Hershcovis's ([23]) model of CWB highlights that frequent and consistent CWB-I (e.g., bullying) have a greater negative impact than infrequent, low-intensity CWB-I. Thus, we expect that exposure to very frequent supervisor CWB-O will drive stronger negative employee reactions, compared to exposure to moderate or infrequent enactment of supervisor CWB-O. In line with past research (e.g., Matta et al., [37]), we hypothesize that these strong negative cognitive and emotional responses will then positively predict workers' intentions to enact CWB. Notably, as AET emphasizes the importance of capturing the characterizing features of events when assessing reactions to those events, we preemptively combine supervisor CWB-O with its contextual features (i.e., frequency, severity, or relationship quality) in our hypotheses (and later vignettes). Thus, we hypothesize and test the effects of differing levels of the overall combination of event feature with event. We have taken this approach in all hypotheses (and subsequent analyses).

Hypothesis 1a: When participants read about infrequent supervisor CWB-O, compared to moderately frequent supervisor CWB-O, they will have lower negative cognitive appraisals.

Hypothesis 1b: When participants read about very frequent supervisor-CWB-O, compared to moderately frequent supervisor CWB-O, they will have higher negative cognitive appraisals.

Hypothesis 1c: When participants read about very frequent supervisor CWB-O, compared to infrequent supervisor CWB-O, they will have higher negative cognitive appraisals.

Hypothesis 2a: Participants' lower negative cognitive appraisals, as a result of infrequent supervisor CWB-O, will predict lower negative emotions, and a subsequent decrease in intentions to enact CWB.

Hypothesis 2b: Participants' higher negative cognitive appraisals, as a result of very frequent supervisor CWB-O, will predict greater negative emotions, and a subsequent increase in intentions to enact CWB.

Another important dimension of supervisor CWB-O is severity. Research indicates that observers differentiate CWB based on the degree of severity (Robinson & Bennett, [47]). Specifically, minor offenses including working slowly and taking excessive breaks are considered low severity forms of CWB-O, while behaviors such as sabotaging projects and theft are considered high severity forms of CWB-O (Hollinger & Clark, [25]; Mangione & Quinn, [36]; Robinson & Bennett, [47]). Notably, Skarlicki and Kulik ([58]) theorize that behavior severity plays an important role in third-parties' interpretations of supervisor mistreatment such that the more severe the mistreatment of a colleague, the stronger the negative reaction by the third-party. Following this, Karelaia and Keck ([29]) found that observers punished leaders less for mild acts of deviance (including CWB-Os such as meeting tardiness, making personal phone calls, etc.) but imposed harsher sanctions on leaders who engaged in more severe behaviors (e.g., withholding important project information). This suggests that exposure to more severe supervisor CWB-O will drive more negative reactions from subordinates, compared to exposure to moderate or less severe forms of supervisor CWB-O. As with the first set of hypotheses and following the emotion-centered model of work behavior (Spector & Fox, [60]), we hypothesize that the stronger negative cognitive and emotional reactions related to exposure to high severity supervisor CWB-O will in turn positively predict subordinates' intentions to enact CWB.

Hypothesis 3a: When participants read about low severity supervisor CWB-O, compared to moderately severe supervisor CWB-O, they will have lower negative cognitive appraisals.

Hypothesis 3b: When participants read about high severity supervisor-CWB-O, compared to moderately severe supervisor CWB-O, they will have higher negative cognitive appraisals.

Hypothesis 3c: When participants read about high severity supervisor CWB-O, compared to low severity supervisor CWB-O, they will have higher negative cognitive appraisals.

Hypothesis 4a: Participants' lower negative cognitive appraisals, as a result of low severity supervisor CWB-O, will predict lower negative emotions, and a subsequent decrease in intentions to enact CWB.

Hypothesis 4b: Participants' higher negative cognitive appraisals, as a result of high severity supervisor CWB-O, will predict greater negative emotions, and a subsequent increase in intentions to enact CWB.

As perceptions of events are inherently shaped by the relational context in which they occur (e.g., Butts et al., [9]; Nishii & Mayer, [42]), our final characteristic involves the quality of the supervisor-subordinate relationship. This is consistent with theory that interpersonal contexts shape others' reactions to an individual's CWB (Reynolds, Shoss, & Jundt, [46]). Further, Cropanzano et al. ([10]) integrated leader-member exchange theory with AET to suggest that the type of relationship subordinates have with their leader will likely impact subordinates' emotional reactions to the leader. They predicted that a more positive relationship with the leader would elicit positive emotions and positive subordinate behaviors, while a negative relationship with the leader will elicit negative emotions and negative subordinate behaviors (see also Dasborough, [11]; Gooty, Connelly, Griffith, & Gupta, [20]). Considering the previous discussion that supervisor CWB-O overall likely drives negative subordinate reactions, we suggest that these negative reactions will be amplified if they occur within the context of a negative supervisor-subordinate relationship. In other words, we argue that negative supervisor behaviors enacted within the context of an already strained, negative relationship will further enhance subordinates' negative feelings (compared to supervisor CWB-O enacted within a neutral or positive supervisor-subordinate relationship context). Notably, according to leader-member exchange theory, when supervisors have high-quality relationships with their subordinates, these relationships are characterized by mutual respect and trust (Graen & Uhl-Bien, [21]; Nishii & Mayer, [42]). This context of trust and respect will likely "buffer" or reduce any potential negative reactions subordinates may have when their supervisor enacts CWB-O, leading to reduced subsequent subordinate intentions to engage in CWB.

Hypothesis 5a: When participants read about supervisor CWB-O in the context of a positive supervisor-subordinate relationship, compared to supervisor CWB-O in the context of a neutral supervisor-subordinate relationship, they will have lower negative cognitive appraisals.

Hypothesis 5b: When participants read about supervisor-CWB-O in the context of a negative supervisor-subordinate relationship, compared to supervisor CWB-O in the context of a neutral supervisor-subordinate relationship, they will have higher negative cognitive appraisals.

Hypothesis 5c: When participants read about supervisor CWB-O in the context of a negative supervisor-subordinate relationship, compared to supervisor CWB-O in the context of a positive supervisor-subordinate relationship, they will have higher negative cognitive appraisals.

Hypothesis 6a: Participants' lower negative cognitive appraisals, as a result of supervisor CWB-O in a positive relationship context, will predict lower negative emotions, and a subsequent decrease in intentions to enact CWB.

Hypothesis 6b: Participants' higher negative cognitive appraisals, as a result of supervisor CWB-O in a negative relationship context, will predict greater negative emotions, and a subsequent increase in intentions to enact CWB.

The current study utilized a within-person vignette design in which each vignette, developed through two preliminary studies, described supervisor enactment of CWB-O. Vignettes were manipulated to reflect high or low frequency of the supervisor's behavior, high or low severity of the behavior, or negative or positive relationship quality. This is in line with AET's emphasis on events, appraisals, emotions, and subsequent behavior occurring at the within-person level of analysis. As mentioned, by manipulating event features we can better understand which characteristics of the event are driving subordinate reactions. Following AET, and the nature of our study design, our ultimate dependent variable is subordinate intentions to enact CWB. Previous research suggests that vignettes can serve as an appropriate method to understand processes underlying intent to engage in a specific behavior (e.g., Aguinis & Bradley, [1]; Alexander & Becker, [3]; Shoss, Jundt, Maurer, & Reynolds, [55]). Additionally, in line with the theory of planned behavior (Ajzen, [2]), intentions to engage in CWB have been shown to significantly predict actual enactment of CWB (Shoss et al., [56]). We briefly summarize the preliminary studies before turning to our main study.

We conducted two preliminary studies to (a) examine salient features of supervisor CWB-O that may impact how subordinate observers respond to these behaviors and (b) develop the vignettes for use in our main study. In the first preliminary study, we interviewed 22 full-time employees, recruited via convenience-sampling procedures. These employees were from a variety of different industries, including health care law, social work, and retail. There were nine male and 13 female participants. The participants' ages ranged from 22 to 64 years old and the average age of the participants was 37.33 years (SD = 14.63). Semi-structured interviews were conducted over the phone and were transcribed by the researcher conducting the interview. Interviewers asked participants to describe the types of CWBs in which their supervisor engaged, how often he/she engaged in these types of behaviors, how these behaviors impacted them (the interviewee), and how they (the interviewee) reacted to said behaviors. Findings indicated a variety of negative cognitive appraisals of observing supervisors enact CWB-O, including disrespect for supervisors, perceptions that it was unfair that supervisors could engage in these behaviors, and lowered subordinate motivation. Participants also reported a variety of negative emotions including frustration, anger, and annoyance. Importantly, the interviews also revealed several conditional characteristics that established the initial consideration of event features that could influence subordinate reactions to supervisor CWB-O. For example, participants clarified that while their supervisor engaged in various CWB-Os, he/she only enacted the behaviors "occasionally" or "rarely" while other participants emphasized that their supervisor engaged in these behaviors "all of the time" or "everyday" – indicating the importance of behavior frequency in how subordinates viewed their supervisors' behavior.

We used the examples from this first preliminary study as the initial foundation of our vignettes. For behavior frequency, we considered both participants' descriptions of frequent or infrequent behavior along with behavior frequency response options used for commonly accepted measures of CWB (e.g., Bennett & Robinson, [6]; Spector, Bauer, & Fox, [59]; Spector et al., [61]). In terms of severity, findings from Robinson and Bennett ([47]) provided quantitative evidence of behaviors considered more or less severe overall. Participant responses then provided more nuanced severity distinctions between behaviors such as taking personal phone calls/using the internet for personal reasons versus coming in late to work/leaving early. Finally, we considered participants' stories which included descriptions of their relationship with their supervisor along with descriptions of good/bad supervisor relationships in the literature (e.g., Dasborough, [11]). Based on this, we created a total of seven vignettes (see Table 1). For consistency, each vignette included a statement about the relationship with the supervisor, the type of the behavior, and the frequency of the behavior. In any given vignette, only one condition was manipulated, and the other two conditions were described using their respective moderate/neutral language (see Table 1). Notably, we used a gender-neutral name (Taylor) for the supervisor in each vignette and avoided using gender-specific pronouns to help reduce potential gender effects.

Table 1. Final study vignettes.

1 The emphasis in each vignette has been added for the reader. Emphases were not included in participants' surveys.

Next, we pilot tested the vignettes to determine that the manipulation was appropriately salient and related to employee reactions. Specifically, qualitative cognitive interviews ensured that participants appropriately noticed variability in behavior severity, behavior frequency, and relationship quality across vignettes, while a pilot survey determined if the vignettes appropriately predicted differences in outcomes. See Appendix A for full details of the pilot testing. Based on these preliminary studies, the vignette language was finalized and is presented in Table 1.

Data were collected from 210 employees contacted via Qualtrics Panels (Qualtrics, [45]). Qualtrics Panels randomly selected participants from stratified regions across the United States who met the study inclusion criteria; in this case, full-time employees who had regular contact with their supervisor (e.g., interacted with their supervisor 3–4 times a week). Qualtrics uses specialized recruitment campaigns to target niche panels. Each sample is proportioned to the general population with qualified participants randomly selected from there (Qualtrics, [45]). Additionally, Qualtrics automatically removes careless responders based on criteria provided by the researchers. In our case, we asked that responders who missed two or more attention checks (out of four) be removed. Using this strategy, there was a 99% chance that a careless responder was appropriately identified (Meade & Craig, [39]). Qualtrics also removed participants who completed the survey too quickly (i.e., one-third the time of the average rate). With these checks in place, Qualtrics provided 210 responses. After further cleaning (i.e., participants whose self-identified industry suggested current unemployment [i.e., stated recently unemployed, or put in miscellaneous words or letters], n = 27; and those with more than 5% missing data, n = 8), we were left with 176 completed surveys with a total within-person N of 1232 cases. Of these participants, the average survey completion time was 27.37 minutes (SD = 24.85)

Our final sample (N = 176) had a representative proportion of males (53%) and females (47%), and a representative proportion of ethnicities (15% Black, 5% Asian, 67% White, and 13% Hispanic). The average age of participants was 40.57 (SD = 12.13). Nineteen percent of the participants had a high school degree, 3% had vocational or technical school training, 32% had some college experience, 29% had a Bachelor's degree, 11% had a Master's degree, and 5% had a Doctoral/Professional degree. Most participants worked between 30–40 hours a week (49%) or 40–50 hours a week (32%), and had an average organizational tenure of 8.24 years (SD = 7.52).

Participants were sent an e-mail by Qualtrics inviting them to participate in the study. Informed consent was obtained from all individual participants included in the study and participation was anonymous. Upon reading the informed consent form, participants completed the survey. Participants read all seven vignettes presented in Table 1, which were randomized to reduce order effects. After reading each vignette, participants answered items about how they appraised the supervisor's behavior in the vignette, their emotional reaction, and the extent to which they might enact CWB considering the vignette they had just read. Items within each scale (i.e., within the appraisal measure, within the emotion measures, and within the CWB measures) were also randomized to reduce issues with order and response effects. After reading all vignettes and reaction questions, participants answered personality items and demographics.

Supervisor CWB-O frequency was captured using three vignettes, representing three-levels of frequency: a vignette that described low-frequency CWB-O (vignette 1: "rarely [only a few times a year]"), a vignette that described moderately frequent CWB-O (vignette 7: "sometimes [once or twice a month]"), and a vignette that described high-frequency CWB-O (vignette 2: "always [everyday]").

Severity of supervisor CWB-O was captured also using three vignettes, again representing three-levels of severity. Vignette 3 presented a situation of low severity ("take personal phone calls, chit-chat rather than work, and use the internet for non-work related things"); vignette 7 presented a situation of moderate severity ("come into work late and leave work early, while also taking longer lunch breaks than everyone else"); and vignette 4 presented a situation reflecting high severity CWB-Os ("falsify financial reports, discuss confidential work information with people outside the organization, and sabotage major projects").

The quality of relationship with the supervisor was manipulated across three vignettes, again reflecting three types of relationships with one's supervisor. Vignette 5 reflected a positive relationship ("In general, your relationship with your supervisor is positive. You believe Taylor cares about your well-being. You'd probably describe your relationship with Taylor as good"); vignette 7 reflected a neutral relationship with the supervisor ("In general, your relationship with your supervisor is fairly neutral and ordinary. You'd probably describe your relationship with Taylor as neither good nor bad"); and vignette 6 described a negative relationship ("In general, your relationship with your supervisor is negative. You believe Taylor makes life at work miserable, and you are, in general, rather discouraged working under Taylor. You'd probably describe your relationship with Taylor as bad").

Cognitive appraisals of each vignette were measured using a six-item scale we developed (using a combination of sources; Dewe & Ng, [12]; Dewe & Trenberth, [13]; Lazarus, [34]; Roseman, Spindel, & Jose, [50]) asking participants' degree of agreement (strongly disagree – 1 to strongly agree – 7) with items reflecting negative cognitive reactions about the supervisor's behavior. High scores on this scale indicated more negative cognitive reactions. Example items include "My supervisor's behavior would negatively impact me and my job" and "I would feel a sense of injustice seeing my supervisor's behavior" (α =.90). The full list of items and evidence of their factor structure are available in Appendix B.

It is important to note our reasoning for developing our measure of cognitive appraisals (rather than using an already existing measure from the literature). Specifically, because of the number of measures and vignettes that participants were reading/completing in our study, we were conscious of participant survey fatigue and effort, and thus wanted a concise measure of cognitive reactions specific to the events that were presented in the vignettes. Notably, Dewe and colleagues (Dewe & Ng, [12]; Dewe & Trenberth, [13]) provide a 23-item primary appraisal measure that was developed based on the specifics of each research study they completed. These items were helpful to give us direction on wording and content of primary/cognitive appraisal items particularly relevant to the workplace, but at 23-items it was too long for the purposes of our study. Roseman et al. ([50]) also included a measure of appraisals that were used to predict emotions, but similar to Dewe and colleagues, the measure was much longer than we needed for the study and was structured for a different type of experimental study design. However, the wording of items again provided insight into typical characteristics that are asked about in an appraisal process. Finally, Lazarus ([34]), serving as a foundational publication in the appraisal/coping literature, provided further example wording and phrasing for possible reactions and coping tactics individuals engage in when presented with a stimulus, particularly a stressful one. Thus, across all these sources we were able to identify commonly used phrasing and wording that had been used across multiple research studies looking at cognitive appraisals and emotions to develop a succinct, while still inclusive, and more directly relevant measure of cognitive appraisals for our study.

To measure participants' negative emotional reactions to supervisor behavior in each vignette, we used the negative affect (NA) items from the short version of the Job-Related Affective Well-Being Scale (JAWS) (Van Katwyk, Fox, Spector, & Kelloway, [62]). Participants indicated the extent to which the supervisor's behavior in each scenario made them feel each emotion ranging from not at all (1) to very much (5). Given the results of the preliminary studies and prior research that separates NA into its high- and low-intensity variants (e.g., Van Katwyk et al., [62]), we examined high- and low-intensity NA separately. High-intensity NA included emotions such as angry and frightened (five items total) whereas low-intensity NA items included discouraged and gloomy (four items total). The scores of both affect scales had acceptable levels of internal consistency (High-intensity NA: α = 92; Low-intensity NA: α =.88).

We used the 10-item version of the CWB-Checklist (Spector et al., [59]) to measure participants' intent to engage in subsequent CWB. Participants were asked to indicate their level of agreement (strongly disagree – 1 to strongly agree – 5) with their inclination to enact CWB after each scenario. Five items assessed CWB-I and five items assessed CWB-O. Example items included "purposely waste your employer's materials/supplies" and "stay at home from work and say you are sick when you aren't." Internal consistency of the scores obtained for both CWB-I and CWB-O was acceptable (α =.89 and α =.84, respectively).

We included positive affect (PA) items (both low- [at ease, relaxed] and high-intensity [ecstatic, inspired] PA) from the JAWS (Van Katwyk et al., [62]) and neutral emotion items (i.e., indifferent, unconcerned, and bored) as within-person level controls. AET (Weiss & Cropanzano, [63]) holds a bipolar conceptualization of affect. It suggests that individuals experiencing a negative mood state are unlikely to experience a positive one. Thus, when mood is assessed as a state, the correlation between NA and PA is expected to be negative and strong. Preliminary study 1 suggested possible theoretical importance of neutral affect. Specifically, we found that individuals often expressed indifference toward their supervisor's CWB if they believed that the behavior would not negatively affect them. Thus, we included PA and neutral affect items to control for them as possible competing mechanisms. Confirmatory factor analyses supported the five emotion-factor structure.[1] Cronbach's alphas indicated acceptable levels of internal consistency for low- (α =.94) and high-intensity (α =.94) PA scale scores. The neutral affect scale scores showed lower internal consistency (α =.58).

Finally, AET classifies employee affective dispositions as contributing factors to emotional reactions and affectively driven behaviors (Weiss & Cropanzano, [63]). Thus, we controlled for neuroticism and extraversion. As participants were already responding to several emotion items specific to each vignette, we decided to use extraversion and neuroticism as controls rather than traditional trait-affect measures. Extraversion is commonly recognized as the five-factor-model personality trait characterized by high PA and energetic tendencies, while neuroticism has been strongly linked to trait tendencies in NA (John, Naumann, & Soto, [26]). Notably, in line with McCrae and Costa's ([38]) five-factor-model of normal personality, we conceptualize neuroticism as the opposite pole of emotional stability and not as a clinical trait (Jones & Arnold, [28]). Each trait was measured using the IPIP-NEO-PI (Johnson, [27]). Internal consistency of the scores obtained for both traits was acceptable (α =.89 for neuroticism and α =.88 for extraversion).

Please note the data that support the findings of this study (and all mentioned preliminary studies) are available from the corresponding author, CK, upon reasonable request. Descriptive statistics and correlations are in Table 2. High- and low-intensity negative emotions were positively related to negative cognitive reactions, CWB-I, and CWB-O. Furthermore, participants' cognitive reactions were also significantly, positively correlated with both types of CWB. These relationships were in the expected directions. Many of the scenarios were also correlated with participant reactions and CWB, providing initial support for the foundation of our model.

Table 2. Descriptive statistics and correlations for all study variables.

• 2 N = 1232 within-person; N = 176 between person. RQ = relationship quality; PA = positive affect; NA = negative affect; Int. = intensity.
• 3 *p <.05; all significance tests are two-tailed.

Intra-class correlations indicated that between-person differences accounted for a large portion of variance in both CWB-I (ICC =.77) and CWB-O (ICC =.69). These ICCs highlight the value of the nested-model and the need to control for between-person characteristics (Bliese, [7]). Thus, we used multilevel modeling in MPLUS (Version 7.4; Muthén & Muthén, [41]) to test our hypotheses. All hypotheses were tested at the within-person level (Level 1) with between-person controls (i.e., neuroticism and extraversion) entered at Level 2. Following recommendations from Hofmann and Gavin ([24]), we also calculated average cognitive appraisals and average affective appraisals across all scenarios (see Figure 2) and entered these in as Level-2 controls. Within-person variables were person-mean centered, and neuroticism and extraversion were grand-mean centered.

Graph: Figure 2. Overall study model with all controls (standardized effect sizes).

All hypotheses were tested using a single model. We created dummy codes for each vignette (six dummy variables total) with vignette 7 (the neutral vignette for each manipulated event feature) coded as 0, and thus, serving as the comparison vignette for all other vignettes. Since vignettes were written such that all event features were represented but only one feature was manipulated in one direction at a time (i.e., vignette 1 only manipulated behavior frequency – specifically low frequency), it was possible to compare coefficients across all vignettes and isolate the effects of a specific event feature. Thus, all vignette dummy codes were entered into the model simultaneously as independent variables (see Figure 1).

We also simultaneously included all affect variables in the model (Level 1) as well as neuroticism and extroversion at Level 2 (see Figure 2). This allowed us to control for any variance accounted for by the PA reactions, allowing us to more clearly parse out the effects of the negative emotional reactions on participants' decisions to enact CWB. All effect-sizes reported in the results section are unstandardized (standardized effect sizes are reported in Figures 2 and 3). Variance component analyses indicated that the within-person variance explained in CWB-I was 65% while the within-person variance explained in CWB-O was 56%. Pseudo-R2 values, calculated as 1 – [(σ2full + τ00full)/(σ2null + τ00null)] (LaHuis, Hartman, Hakoyama, & Clark, [31]), were.27 for CWB-O and.30 for CWB-I.

Graph: Figure 3. Study model with only hypothesized paths. Standardized effect sizes are the same as in Figure 2 and account for control variables; control variables have been visually removed to improve visibility of hypothesized relationships.

Hypotheses 1 and 2 focused on the impact of frequency of supervisor CWB-O on participants' cognitive and affective reactions, and subsequent decision to enact CWB. As seen in Figure 3, Hypothesis 1a is supported. Specifically, participants had reduced negative cognitive appraisals after reading the low-frequency CWB-O vignette (V1; b = −.62 (.10), p <.05) compared to when they read the moderately frequent CWB-O vignette (V7). Hypothesis 1b is also supported such that participants had greater negative cognitive appraisals after reading the very frequent supervisor CWB-O vignette (V2; b =.81 (.10), p <.05), compared to when they read the moderately frequent supervisor CWB-O vignette (V7). Results also showed support for Hypothesis 1c. Following recommendations from Lau and Cheung ([33]), we calculated a difference score between the vignette 1 (low frequency) versus vignette 7 (moderate frequency) effect and the vignette 2 (high frequency) versus vignette 7 (moderate frequency) effect to determine whether there was a difference between reading about infrequent versus very frequent supervisor CWB-O on cognitive reactions. These results indicated that there is a significant difference between very frequent and infrequent CWB-O (difference = 1.49 (.14), p <.05), such that reading about very frequent behaviors had a stronger, positive effect on negative cognitive appraisals.

Next, Hypothesis 2a is supported such that reading about infrequent supervisor CWB-O had significant, negative indirect effects on participant high- (b = −.38 (.06), p <.05) and low-intensity (b = −.28 (.05), p <.05) negative affect (NA). Infrequent supervisor CWB-O (V1) also had significant, negative indirect effects on participant intentions to enact both CWB-I (b = −.05 (.01), p =.001) and CWB-O (b = −.10 (.02), p <.05). Total indirect effect analyses indicated that the indirect effect of infrequent supervisor CWB-O on participant intentions to enact CWBs functioned as a result of reduced negative cognitive reactions driving reduced negative emotions (see Table 3). Hypothesis 2b is also supported such that reading about frequent supervisor CWB-O resulted in significant, positive indirect effects on high- (b =.45 (.05), p <.05) and low-intensity (b =.33 (.04), p <.05) negative emotions, as well as participants' decisions to enact CWB-I (b =.06 (.01), p <.05) and CWB-O (b =.12 (.02), p <.05). Results indicated that these positive effects on participant CWB developed as a result of participants' greater negative cognitive appraisals driving greater negative emotions (see Table 3).

Table 3. All nested, indirect effects.

• 4 N = 1232 ratings nested within 176 raters. Bolded values are significant
• 5 V# = vignette number (see Table 1), CogReact = cognitive reactions, PA = positive affect, NA = negative affect.
• 6 *p <.05.

Hypotheses 3 and 4 predicted that severity of supervisor CWB-O would have a significant impact on participants' negative reactions, and subsequent intentions to enact CWB. As seen in Figure 3 and Table 3, Hypothesis 3a is not supported. Specifically, there was no significant effect on participants' cognitive appraisals after reading the low severity supervisor CWB-O vignette (V3) compared to reading the moderately severe supervisor CWB-O vignette (V7). However, Hypothesis 3b is supported such that when participants read the high severity supervisor CWB-O vignette (V4; compared to reading V7 – moderately severe supervisor CWB-O), they reported greater cognitive reactions (b = 1.70 (.13), p <.05). Hypothesis 3c is also supported. We again calculated a difference score between the effect of vignette 3 (low severity) versus vignette 7 (moderate severity) and the effect of vignette 4 (high severity) versus vignette 7 (moderate severity) to determine whether there was a difference between low severity and high severity supervisor CWB-O on negative cognitive reactions. Results indicated that there is a significant difference between high severity and low severity supervisor CWB-O (difference = 1.82 (.14), p <.05), such that reading about high severity supervisor CWB-O had a stronger, positive effect on negative cognitive appraisals.

Hypothesis 4, then, predicted the indirect effects of low severity and high severity supervisor CWB-O on participants' negative emotions and intentions to enact CWB.

Hypothesis 4a is not supported as there were no significant indirect effects on participants' negative emotions or decisions to enact CWB when participants read the low severity vignette. Hypothesis 4b is supported, however, as reading about high severity supervisor CWB-O resulted in significant, positive indirect effects on participants' negative emotions (high-intensity: b =.93 (.08), p <.05; low-intensity: b =.70 (.07), p <.05), as well as participants' decisions to enact CWB-I (b =.12 (.03), p <.05) and CWB-O (b =.24 (.04), p <.05). These indirect effects functioned as a result of participants' greater negative cognitive reactions positively predicting negative emotions (see Table 3).

Finally, Hypotheses 5 and 6 predicted that relationship quality with a supervisor would influence participants' reactions to supervisor CWB-O. As seen in Figure 3, Hypothesis 5a is not supported, in that there was not a significant effect on participants' cognitive reactions to supervisor CWB-O when they read about it in the context of having a positive relationship with the supervisor (V5) versus a neutral relationship (V7). However, findings did indicate support for Hypothesis 5b. When participants read about supervisor CWB-O in the context of a negative relationship with the supervisor (V6; compared to a neutral relationship, V7), they had greater negative cognitive reactions (b = 1.32 (.12), p <.05). Findings also indicated support for Hypothesis 5c as the difference score comparison of the effect of vignette 6 (negative relationship) versus vignette 7 (neutral relationship) and the effect of vignette 5 (positive relationship) versus vignette 7 (neutral relationship) was significant (difference = 1.46 (.13), p <.05). Specifically, participants had a greater negative cognitive reaction after reading about CWB-O enacted within a negative supervisor-subordinate relationship context compared to reading about CWB-O enacted within a positive supervisor-subordinate relationship context.

With respect to Hypothesis 6 and the indirect effects of supervisor CWB-O enacted under different types of supervisor-subordinate relationships, there were no significant indirect effects on participants' negative emotions or decision to enact CWB when reading about a positive relationship (V5) with the supervisor (see Table 3), suggesting Hypothesis 6a is not supported. However, and in support of Hypothesis 6b, supervisor CWB-O enacted within the context of a negative supervisor-subordinate relationship had significant, positive indirect effects on high- and low-intensity negative emotions (b =.73 (.07), p <.05; b =.54 (.06), p <.05, respectively), as well as participants' decisions to enact CWB-I (b =.09 (.02), p <.05) and CWB-O (b =.19 (.03), p <.05). As with prior hypotheses, these indirect effects were a function of greater negative cognitive reactions, and greater negative emotions (Table 3).

Overall, the present study continues the research stream highlighting the profound influence supervisors have over subordinates' work experience (e.g., Cropanzano et al., [10]; Shoss et al., [54]); in this case, through their enactment of CWB-O. When a supervisor enacts behaviors that go against subordinates' expectations, such as CWB-O, they elicit negative appraisals and negative emotions from subordinates, in turn increasing subordinates' intentions to enact CWB (e.g., Dasborough, [11]; Rupprecht et al., [51]). While much of the current research in this domain has predominantly focused on observing supervisor CWB-I (i.e., abusive supervision; Hershcovis, [23]), this study advances theory and research by contributing a foundational framework for the influence and consequences of the vicarious impact of supervisor CWB-O. Our findings suggest that subordinate reactions are influenced by behavior frequency and severity, and quality of the supervisor-subordinate relationship.

Specifically, supervisor CWB-O frequency significantly predicted subordinate cognitive and emotional reactions, which, in line with AET (Weiss & Cropanzano, [63]), significantly influenced subordinate intentions to enact CWB themselves. Both high and low levels of frequency influenced subordinate intentions. Participants reported decreased intentions to enact CWB when the supervisor in the vignette engaged in less frequent CWB-O, through decreased negative cognitive and emotional reactions. When the supervisor in the vignette enacted CWB-O more frequently, participants reported increased negative cognitive appraisals and emotional reactions, driving increased intentions to enact CWB. This supports previous theory and research which suggests that more frequent CWB (i.e., consistent and persistent negative behaviors) have stronger negative impacts on employees (Hershcovis, [23]). Findings also support theories which suggest that negative events are often not singular, and negative reactions can build on each other (Frijda, [19]; Zillmann, [65]), eventually driving increased negative emotions and emotion-driven behavior such as CWB (Spector & Fox, [60]).

Next, results indicated that severity of supervisor CWB-O on intentions to enact CWB was partially supported. Specifically, more severe events of supervisor CWB-O increased intentions to engage in CWB through increased negative emotional reactions driven by increased negative cognitive appraisals. The low severity condition did not significantly predict cognitive or emotional reactions and did not have an indirect effect on participant intentions to enact CWB. This lack of reaction to smaller caliber CWB-O could be the result of an overall norm or understanding that minor CWB-O are generally acceptable behaviors (Shoss et al., [56]). It is also possible that subordinates view supervisors as fallible and may give them a pass for less severe behavior. However, increased intentions to engage in CWB following severe supervisor CWB-O could be subordinates' reactive coping mechanism intended to restore an injustice felt toward their supervisor (Folger & Skarlicki, [17]; Skarlicki & Folger, [57]). In other words, the impact of severe supervisor CWB-O on subordinate reactions supports justice theory which suggests that behavior severity plays a large role in subordinates' interpretations of witnessing or learning about supervisor mistreatment (Skarlicki & Kulik, [58]). The stronger negative reactions to the more severe supervisor CWB-O suggest that subordinates interpret these behaviors as greater mistreatment or injustice that needs to be addressed. Notably, high severity behaviors (e.g., theft, confidentiality breaches) can be difficult for others to observe, but examples of these types of supervisors' behaviors were gathered in our qualitative preliminary data (e.g., falsifying receipts) and accusations can be found in the news (e.g., in 2018, the CEO of WPP, the world's largest advertising agency holding company, resigned after staff claims that he had misused company assets; Dishman, [14]), suggesting employees do witness even these more severe forms of CWB-O.

Finally, negative relationship quality was significantly related to participant intentions to enact CWB through increased negative cognitive appraisals and subsequent negative emotional reactions. It is possible that subordinate intentions to enact CWB may again be a reaction of retaliation or resentment (e.g., Skarlicki & Folger, [57]) against a supervisor who fosters negative working relationships. Positive subordinate–supervisor relationship quality did not significantly predict cognitive or emotional reactions or have any indirect effect on subordinate intentions to engage in CWB. It may be that a positive relationship with a supervisor acts as a buffer resulting in an indifferent reaction to supervisor CWB-O (as opposed to the hypothesized decrease in negative reactions). Subordinates may feel it is more acceptable to let some behavior slide when they have otherwise positive experiences with their supervisor (e.g., Bradfield & Aquino, [8]). The significant impact of supervisor CWB-O on participant cognitions and emotions within a negative relationship context supports previous research that negative exchanges with supervisors serve as significant negative emotion-eliciting events for subordinates (Gooty et al., [20]).

Our findings confirm that supervisors' organizationally deviant behavior does influence subordinates' attitudes and behaviors even if not targeted directly toward an individual. This advances research on the impact of vicarious CWB (Robinson et al., [48]) and suggests that even when no particular person is being hurt by the CWB, employees can still develop negative reactions to these behaviors, driving potential enactment of their own CWB. Furthermore, by clearly defining the conditions of influence supervisors have in shaping subordinates' work experience, the data offer further insight into understanding what drives subordinates' CWB. Specifically, high severity supervisor CWB-O and negative relationship quality had the strongest impacts on subordinates' cognitive and emotional reactions and subsequent decisions to enact CWB. These findings can help guide research and practice in continued efforts to understand why CWB persist in the workplace.

The current results suggest that beyond the organizational impact a supervisor's CWB-O may have, these behaviors can have ripple effects throughout the organization through their impact on the supervisor's subordinates. Thus, organizations may be well served to look not only at the organizational impacts of CWB-O, but also the negative individual impacts these behaviors can have on subordinates. Supervisor CWB-O may not be considered the first place to look as a source of influence for subordinate CWB; however, our results suggest that this is a possibility.

Additionally, our findings can offer insight into how to best improve practices or adjust behavioral norms and expectations. Specifically, we identified three key features of supervisor CWB-O (frequency, severity, and relationship quality) that should be considered when determining the impact of supervisor CWB-O. While in general CWB-O are likely undesirable for an organization, these behaviors may be less concerning depending on how frequently they are occurring, how severe they are, and the types of relationships the supervisor has with his/her subordinates. If supervisors are engaging in CWB-O frequently, they may be establishing company norms that these behaviors are acceptable, at least for leadership. This can foster negative perceptions and emotions among subordinates, resulting in a negative work environment. The same concerns exist if supervisors are engaging in particularly severe forms of these behaviors, and/or have negative relationships with their subordinates. These would likely be concerns that top leadership would need to address. However, there may be less of a need to target resources aimed at reducing supervisor CWB-O (at least as a main source of negative norms and negative subordinate experiences driving intentions to engage in CWB) if further investigation suggests that supervisors engage in these behaviors infrequently, engage in less severe forms, and/or have a positive relationship with their subordinates otherwise. Thus, we provide a guiding framework to help organizations identify under what conditions supervisor CWB-O may be most impactful on subordinates.

We believe that the multi-study approach to developing the vignettes and identifying the overall hypothesized model is a particular strength of our study. We used both qualitative and quantitative as well as deductive and inductive approaches to develop as realistic vignettes as possible. Additionally, our overall model is grounded in and continues to support a foundational organizational theory: AET. However, our study is not without its limitations.

We note that our final model data were collected in a single time-point, suggesting possible concern of common-method bias (Podsakoff, MacKenzie, & Podsakoff, [43]). Nested confirmatory factor analyses suggest support for the study's eight-factor model (consisting of five emotion factors, two CWB factors, and one cognitive factor) and that this model has a better fit than a single-factor model.[2] Additionally, vignettes and items were randomized so that participants received the conditions and items in different orders to reduce the possibility that response patterns and order effects might impact the relationships.

A second limitation worth noting is that we assessed intentions to enact CWB rather than the actual enactment of CWB. Notably, intentions to engage in CWB have been found as a significant predictor of the actual behavior (Shoss et al., [56]), and intentions overall serve as a significant predictor of most behaviors (e.g., Ajzen, [2]). While our study did not assess subordinate behavior specifically, our vignette-approach did allow for a differentiated look at the potential decision-making processes underlying subordinate reactions to supervisor CWB-O (following Shoss et al., [55]). We manipulated key aspects of each vignette to specifically identify which event characteristics have the strongest impact on subordinate reactions to supervisor CWB-O that subsequently result in decisions to enact CWB. However, it would be valuable to replicate these findings using other methodologies. Example methodologies include retroactive recall where participants are asked to share an occurrence of supervisor CWB-O and describe their reactions, or an experience-sampling methodology where the specific occurrences of supervisor CWB-O can be tracked and subordinates' reactions to these behaviors documented.

A third limitation is that we used an online source for our sample which could raise questions about our sample representativeness and its impact on the external validity of our findings (e.g., Landers & Behrend, [32]). However, Qualtrics Panels may be a stronger sample source than other online sampling methods (e.g., MTurk; convenience sampling through social media sites, etc.) as Qualtrics uses a stratified, random sampling approach to ensure a sample representative of the adult working population. They also use an authentication system to ensure the participant is legitimate (Qualtrics, [45]). However, it is also important to note that research has shown that organizational/group norms may play a significant role in CWB within an organization (e.g., Ferguson & Barry, [16]). Thus, it would be worthwhile to replicate this study in the field, and account for the potential effects of organizational norms and other organizational nuances on subordinate reactions to supervisor CWB-O.

Finally, future research should also consider other characteristics that could influence subordinate reactions to supervisor CWB as well as the potential for interacting effects of the event characteristics studied here. For example, would infrequent yet severe CWB-O have a greater impact on subordinates' reactions than frequent but less severe behaviors? Would subordinates still be willing to "look the other way" for a supervisor enacting severe forms of CWB-O if they have a positive relationship with the supervisor? As the vignettes in the current study were written purposefully to isolate the effects of each event feature, we were not able to test possible combined effects. However, combinations of event characteristics would be worth studying. Additionally, other moderators, such as subordinate characteristics (e.g., moral standards; Klotz & Bolino, [30]), opportunities for supervisors and/or subordinates to enact CWB, and/or organizational norms and policies (Shoss et al., [56]) could be studied.

Overall, the current findings suggest that even though behaviors may not be targeted at any particular individual, they can still have a significant impact on others in the workplace. If CWB-O are carried out by those who significantly shape the subordinate experience (i.e., the supervisor), and if the behaviors are frequent, severe, or within the context of a negative supervisor-subordinate relationship, they may significantly, negatively impact subordinates' cognitions and emotions. These negative appraisals may result in increased subordinate intentions to engage in their own CWB. This suggests a ripple effect of CWB throughout an organization, driving increased negative reactions and negative behaviors.

We would like to thank Alyssa De Santi and Debarati Majumdar for their gracious help with data management during the second pilot study and final study.

The authors declare that they have no conflict of interest.

Once we established the low (i.e., infrequent, minor, or positive relationship quality), moderate/neutral, and high (i.e., frequent, severe, or negative relationship quality) levels of each condition, we combined these characteristics together into a total of seven vignettes. Once the vignettes were written, we conducted our pilot study to determine whether the conditions in each vignette were appropriately manipulated. Specifically, we sought to determine whether participants would notice which detail in each vignette was manipulated. For example, in the severity vignettes, participants should specifically identify behavior severity (as opposed to behavior frequency or supervisor-subordinate relationship quality) as the most important characteristic of the vignette, and then whether the behavior seemed severe or not. Finally, we also expected significant differences between the high and low levels of each condition in participants' appraisals of supervisor behavior and their emotional reactions (according to AET; Weiss & Cropanzano, [63]). To answer these questions, we collected both qualitative and quantitative data.

For the qualitative portion of the pilot study, and to test the manipulation check of each vignette, 12 undergraduates at a Midwestern university who worked part-time were recruited to participate in one-on-one cognitive interviews about the vignettes (Willis, 1999). These participants were interviewed by the second author who received training on cognitive interviewing prior to the study. In each interview, participants read each vignette and looked through the cognitive appraisal and emotion items used in the main study. After reading a vignette, participants described to the interviewer their thought-process as they read the vignette, the specific details that stood out to them, and how these details influenced their reactions (Willis, 1999). The interviewer asked the participants to "Tell me what you were thinking as you read through this scenario" and followed up with example probes such as "What to you are the key details in this scenario" and "How did these detail(s) influence your emotional reaction or appraisal of the scenario." Each interview was audio-recorded, and later transcribed and coded by three coders. Specifically, each interview was coded to determine which detail (i.e., behavior frequency, relationship quality, or behavior severity) in each vignette was described as most salient to the participant. This served as the manipulation check to determine that participants noticed that our key variables – severity, frequency, and relationship – varied across vignettes. All interviews were coded by the same three coders and were coded separately for a total of 252 codes (based on 12 participants rating 7 vignettes with 3 elements manipulated across the 7 vignettes). Raw agreement indices indicated the coders agreed about which detail was most important to a participant in a given vignette 93% of the time. For vignettes where there was disagreement, all three coders discussed the specific vignette and resolved any differences.

Once there was agreement on detail saliency for each vignette, the coders then determined whether the participants correctly identified the manipulated characteristic (i.e., for the severity condition, did participants identify the severe behaviors as the most important information in the vignette). Across these codes, participants correctly identified the manipulated detail for the low and high severity vignette, and the negative relationship vignette. For the positive relationship vignette, most participants identified the relationship as the salient detail, but only a slight majority (63% of the participants). Additionally, for the low- and high-frequency vignettes, participants did not identify frequency as the manipulated characteristic. This parallels findings in the repeated measure analyses described below where high-frequency did not significantly differ from moderate frequency in impacting outcomes.

To further refine our vignettes and pilot test their effectiveness in predicting differing levels of participants' cognitive and emotional reactions to supervisor CWB-O, data were collected from 200 participants from Amazon's MTurk. After removing incomplete surveys and careless responders (i.e., participants who missed two or more attention check questions out of four possible attention checks), we had within-person data from 181 participants. In the pilot survey, participants read each vignette (order randomized), and each vignette was then followed by the same cognitive appraisal items and emotion items used in the main study. As with the main study, the neutral and PA measures were included to prevent hypothesis-guessing. Additionally, participants completed the neuroticism and extraversion measures from the IPIP-NEO-PI (Johnson, [27]) which served as between-person covariates in the repeated measure analyses.

According to the repeated measure analyses, there were significant differences between the low and high levels for each condition (severity, frequency, and relationship quality) for each outcome (i.e., participant appraisals, high NA, low NA, high PA, low PA, and neutral affect). Notably, there were no significant differences between the high-frequency vignette and its moderate counterpart on all affective-outcomes. This confirmed the qualitative analyses that suggested the differences in frequency between vignettes were not strong enough for participants to notice the manipulation; thus, the variability in the manipulated characteristic was not strong enough to elicit differing reactions from participants.

Combined, these pilot findings suggested that the high-/low-frequency vignettes needed to be re-written to improve saliency of the manipulated characteristic. Finally, since it was not definitive that participants appropriately identified the positive relationship as a manipulated characteristic in the qualitative pilot study, we re-wrote the positive relationship vignette. The final vignettes are in Table 1.

Willis, G. B. (1999). Cognitive interviewing: A "how to" guide. In R. A. Caspar, J. T. Lessler, & G. B. Willis (Chairs) short course Reducing survey error through research on the cognitive and decision process in surveys presented at the 1999 Meeting of the American Statistical Association.

(based on wording from Dewe & Ng, [12]; Dewe & Trenberth, [13]; Lazarus, [34]; Roseman et al., [50])

• My supervisor's behavior would negatively impact me and my job.
• My supervisor's behavior would be stressful.
• I would consider my supervisor to be incompetent.
• I would be able to cope with my supervisor's behavior. (R)
• I would feel a sense of injustice seeing my supervisor's behavior.
• I feel my supervisor's behavior was inappropriate.

Confirmatory factor analyses suggested support for the unidimensionality of this measure. More specifically, fit indices indicated acceptable fit: χ2 = 1362.10 (15), CFI =.93, RMSEA =.09 (.08,.11), SRMR =.04, and all items had acceptable standardized factor loadings (>.4; five of the six items with factor loadings >.78). The item with the lower loading is item #4 which was the reverse-scored item, and thus likely has a lower loading partially as a result of difference in framing. The Cronbach's alpha for this scale's score is.90.