Bystander Intervention in the Context of Abusive Supervision: Effects of Power Distance Orientation and Gender

Introduction

In work contexts, aggressive and unethical actions of powerful managers may have a negative impact on all employees. A coworker may be the victim of an abusive supervisor, and a bystander may sense or witness the problem without hearing the victim say anything about it. Observers may also feel guilty simply for being an outsider in these situations whereby they are aware that someone in their environment is suffering (Mitchell et al., 2015; Priesemuth, 2013). In the event of abusive supervision, which individual and situational factors make third-party observers more likely to speak up and intervene and, if they do decide to intervene, are they more likely to provide direct or indirect support? This study, based on the premises of bystander intervention framework (Latané & Darley, 1970) investigates these questions.

Examples of abusive supervision include criticizing, ridiculing, intimidating, humiliating subordinates in front of others, making negative comments about them to others, or being rude to them (Tepper, 2000). It is possible that such manners and behaviors have negative effects on the victims, the observers, and the organizations, as well as causing a significant risk towards the sustainability of a psychologically healthy work environment (Martinko et al., 2013; Martinko et al., 2012; Tepper, 2007). Previous research has primarily focused on the precursors of abusive supervision and its effects on the victims (Lee et al., 2018; Martinko et al., 2013; Tepper, 2007; Tepper et al., 2011; Zhang & Bednall, 2016). However, the mechanisms guiding the perceptions of third-party observers have not been studied adequately, and observers’ reactions, as governed by their perceptions may, in fact, be crucial factors for discouraging such unethical behaviors in the work context (D’Cruz & Noronha, 2011; van Heugten, 2011).

The perception of observers may be governed by several contextual factors such as the online vs. offline nature of bullying or bystanders’ relationship with the victim (Coyne et al., 2019) and the ethical climate prevalent in the organization (Bulutlar & Oz, 2009). These perceptions are important for three main reasons. First, though observers are not the targets of the abuse, they witness the abuse as well as its negative effects on the victims. It is possible that witnessing such situations may lead to feelings of vicarious abusive supervision, which may then impact observers negatively. These negative effects mainly depend on how observers perceive the situation (Haggard et al., 2011; Harris et al., 2013). Second, how observers perceive the situation may govern their decision-making process, which in turn will lead them towards supporting (or not supporting) the victim. Support from an observer may reduce the loneliness of victims and also encourage the victim to take part against the perpetrators. In addition, observers’ voices may have a deterrent effect on perpetrators by showing them that victims are not alone. Third, caring for others, i.e., having concern for the well-being of others, is a component of ethical climate (Martin & Cullen, 2006), and observers may impact this climate positively via their caring behaviors targeting the victims, which are likely to be guided by their perceptions. In short, observers have the capacity to serve as change agents (van Heugten, 2011) and contribute to the tackling of abusive behaviors in work settings (D’Cruz & Noronha, 2011).

This study examines observers’ decision-making processes in the event of abusive supervision based on the premises of the Bystander Intervention Framework (Latané & Darley, 1970). The five stages of the framework were used to conceptualize the steps of the decision-making process observers go through in their ways to decide whether to help (or not help) victims of abusive supervision in their work contexts. Each stage will be explained in detail with regard to relevant study hypotheses. Specifically, it was posited that after the initial noticing stage, the second stage – interpretation (i.e., perceived acceptability of abusive supervision) – is influenced by three factors: (i) observers’ power distance orientation, (ii) situational factors such as gender of the perpetrator and the observer, and (iii) interaction between these factors. Power distance orientation of observers is predicted to be a crucial factor, as abusive supervision is a condition that arises from the exertion of legitimate power (French & Raven, 1959) held by supervisors in an unethical way. Thus, abusive supervision emerges as a power issue (Tepper, 2007; Tepper et al., 2009), whereby perpetrators may use abusive supervision as a conduit to gain greater status (Hu & Liu, 2017). As a result, the conceptualization of power by observers is a significant factor which can be examined through observers’ power distance orientations (Lian et al., 2012; Lin et al., 2013; Wang et al., 2012). It is also argued that gender is an important factor which accompanies power distance orientation in the perception of injustice situations (Lee et al., 2000).

The few previous studies on bystanders’ perceptions and role in abusive supervision have either focused on the sole effects of personal values such as power distance orientation (Lian et al., 2012; Lin et al., 2013; Wang et al., 2012) or gender variables such as the gender of the victim (Ouyang et al., 2015; Wang et al., 2016). However, given the complexity of interactions between personal and situational variables in real life situations, the Bystander Intervention Framework (Latané & Darley, 1970) in the abusive supervision context should be studied through methods which allow for assessment of these relationships. With this purpose, this study was conducted using an experimental design allowing for the control of the effects of two key gender-related variables: gender of the perpetrator and gender of the victim. Any other situational variables (e.g., specific behaviors of the perpetrator, coworkers of the victim, organizational context, etc.) were kept constant for each observer. Given the paucity of research focusing on interrelated roles of power distance orientation and gender variables in the perceptions and reactions of observers in abusive supervision situations, the results of this study are expected to contribute to the literature by providing a theoretical explanation for the direct/indirect action versus inaction of the observers.

Stage 1: Noticing the Situation

The first stage in the Bystander Intervention Framework is defined as the phase of noticing an unusual situation. In this phase, observers witness a situation and recognize there is a problem, which may cause harm to the victim (Latané & Darley, 1970). Abusive supervision is defined and measured as a perception, meaning the subordinate’s (or victim’s) perception of the situation is used as the basis for labeling the situation as an abusive supervision case (Martinko et al., 2013; Tepper, 2000). Similarly, the perception of observers is critical for the labeling of the situation. In line with the first expected stage of the bystander intervention model, it was expected that participants would notice that the superior in the case was abusing his/her subordinate.

Hypothesis 1: The majority of participants will notice that the supervisor is abusing her/his subordinate.

Stage 2: Interpretation of the Situation

After noticing the situation, in the second stage of the Bystander Intervention Framework (Latané & Darley, 1970) it is predicted that observers will interpret the situation and decide if it is a situation which needs intervention. In this model, if the situation is perceived as an emergency, it is predicted that bystanders are more likely to provide help. Previous research revealed that there are varying factors which influence observers’ perception of risky situations. These factors can be personal, such as observers’ moral identity (Mitchell et al., 2015) and attribution styles (Martinko et al., 2012), or situational, such as the presence of other observers (“diffusion of responsibility”; Darley & Latané, 1968). Two factors appear to be critical in the abusive supervision context: observers’ power distance orientation (as a personal factor) and the match between genders of observer and victim (as a situational factor). It is predicted that these factors will impact observers’ perception and interpretation of the situation in relation to each other.

Previous studies have shown that the degree of power distance orientation in a given culture impacts perception of abusive supervision. For example, abusive supervision is perceived to be less acceptable in low power contexts, such as Australia (Khan, 2014) and the USA (Vogel et al., 2015), compared to high power distance contexts, such as Pakistan (Khan, 2014), Singapore and Taiwan (Vogel et al., 2015), and China (Pan & Lin, 2017). Similarly, people with high power distance orientation at individual level are expected to respect authority and be less likely to question it (Earley & Erez, 1997; Maznevski et al., 2002). Abusive supervision is a specific case which reflects a power-based relationship between supervisor and subordinate. Specifically, a supervisor’s tendency to abuse his/her subordinates is grounded in his/her power over subordinates. Thus, abusive supervision serves as a sign of power play in an unethical way (Tepper, 2007; Tepper et al., 2009).

Previous studies on abusive supervision revealed that subordinates’ power distance orientation impacts their perception of abusive supervision so that high power distance subordinates were less likely to perceive the situation to be unfair (Lian et al., 2012; Wang et al., 2012) and they were also less likely to suffer from the negative effects of abusive supervision on them (Lin et al., 2013). Given that the victim’s power distance orientation may impact her/his perception of abusive supervision, observers’ perceptions could also be influenced by their own power distance orientations. Thus, observers with a high power distance orientation would be less likely to question the situation and be more likely to perceive the situation as an acceptable supervisor-subordinate relationship.

Hypothesis 2: Observers with a higher power distance orientation are more likely to perceive the abusive supervision situation as acceptable in comparison to observers with a lower power distance orientation.

Another predicted key issue in the perception of abusive supervision is gender. Gender, in relation to power distance orientation, is found to be a main determinant of perception of organizational situations which entail injustice against a certain person (Lee et al., 2000), such as bullying (Salin, 2011). The Social Identity Theory (Tajfel, 1974) states that people automatically categorize others as in- vs. out-group members based on perceived similarities or differences. This social categorization is followed by identification with in-group members, which results in feeling closer and being more tolerant toward them. The judgment of a certain type of behavior should therefore be examined in relation to the characteristics of the person displaying this behavior. It is predicted that the impact of power distance orientation on perceived acceptability of abusive supervision will depend on the match between genders of perpetrator and observer, so that perceived similarity to the power figure may strengthen the relationship as a result of the social identification process (Tajfel, 1974). Due to the perceived similarity to the power figure, observers would be less likely to question the negative behavior of this person, so that the effect of power distance orientation is expected to be stronger for male observer-male perpetrator and female observer-female perpetrator dyads, in comparison to dyads where the genders do not match.

Hypothesis 3: The relationship between power distance orientation and perceived acceptability of abusive supervision is stronger when observer and perpetrator are of the same gender.

Stage 3: Taking Responsibility

In the third phase of the Bystander Intervention Framework (Latané & Darley, 1970), observers who interpret the situation as needing intervention must decide if they can or should take the responsibility for intervening. The Theory of Planned Behavior (Ajzen, 1991) states that attitudes predict intentions, and then those intentions predict actual behaviors. In the abusive supervision context, perceived acceptability of the situation may be the key attitude which impacts observers’ willingness to help the victim either directly, by taking high personal risk, or indirectly, through taking minimum personal risk. Direct help requires active involvement of observers by, for example, informing the management or warning the perpetrator. Indirect help is often passive, meaning observers’ do not need to shoulder the main responsibility. Encouraging the victim to take action is an example of indirect help (Darley & Batson, 1973; Pearce & Amato, 1980). In particular, higher levels of acceptability often decrease the desire for an observer to provide either direct or indirect help, and increase the desire to avoid interfering with the situation.

Hypothesis 4: Higher perceived acceptability (a) increases the likelihood of avoiding to support the victim, (b) decreases the likelihood of providing indirect support, and (c) decreases the likelihood of providing direct support to the victim.

In this study, regarding the perception of abusive behaviors, the gender of the perpetrator displaying such behaviors was examined; it was predicted that an observer-perpetrator gender match would be a significant factor (Hypothesis 3). In addition, it was predicted that the gender of the victim would be a determinant of observers’ willingness to take action against the abusive supervision based on the Gender Schema Theory (Bem, 1981). This theory posits that women are perceived to be weaker than men, thus they need help, whereas men are perceived to be strong and tough and able to solve their own problems. The concept of benevolent sexism, included in the Ambivalent Sexism Theory (Glick & Fiske, 1996), also states that both men and women perceive women to be more vulnerable and prefer to provide more help to them compared to providing help to men, so that, for example, people are more likely to believe that women should be rescued before men in a disaster. Previous literature also shows that women receive more help than men (Eagly & Crowley, 1986). A victim’s gender was also found to be a strong predictor of helping behavior in bullying and mobbing, whereby female victims were more likely to receive help from observers (Salin, 2011). Therefore, in abusive supervision cases, observers’ tendencies toward avoiding intervention are predicted to be higher for male victims, compared to female victims.

Hypothesis 5: Observers’ tendencies to avoid providing support to the victim are higher for male victims, compared to female victims.

Stage 4: Deciding How to Help

In the final stages of the Bystander Intervention Framework (Latané & Darley, 1970), observers who feel they can take responsibility decide on the type of help they can provide. Types of help can be conceptualized as direct or indirect (Darley & Batson, 1973; Pearce & Amato, 1980). As a further consequence of benevolent sexism (Glick & Fiske, 1997), observers’ preference to provide direct or indirect help may be influenced by gender of the victim. Observers may be more likely to take a risk if the victim is female, as females are generally perceived as being weaker, and this means that providing direct help entails a higher risk for the bystander. The type of help which is often afforded to male victims is primarily indirect, as males are generally viewed as having the courage and power to take care of themselves, meaning that observers are more likely to take a risk when the victim is female. In summary, it is expected that the impact of perceived acceptability on tendencies to help is dependent on the gender of the victim, meaning that female victims are more likely to receive direct help and male victims are more likely to receive indirect help.

Hypothesis 6: The effect of perceived acceptability (a) on direct help will be stronger for female victims, (b) whereas its effect on indirect help will be stronger for male victims.

Stage 5: Providing Help

In the original framework, Latané and Darley (1970) argue that observers who reach stage four and decide on the type of help they can provide are quite likely to get involved and provide help to the victim in the situation. This assumption is also consistent with the Theory of Planned Behavior (Ajzen, 1991), which states that intentions are strong predictors of actual behaviors. Therefore, although actual behaviors in real-life settings were beyond the scope of the present study due to the experimental design, intentions prove to be strong indicators of actual helping behavior.

Method

Design and Procedure

The experimental procedure took one week, where participants received a total of six e-mails from Monday until Saturday, with the final requirements which had to be completed in two days (due the following Monday morning). Data collection took place over a total of one year. During the period of data collection, each week started with a maximum of 20 participants. On Monday, first participants signed the informed consent form, responded to demographics questions, and filled out the power distance orientation scale. Afterward, each participant was randomly assigned to one of the dyadic conditions: (i) female perpetrator-female victim, (ii) female perpetrator-male victim, (iii) male perpetrator-female victim, (iv) male perpetrator-male victim. Participants read a paragraph introducing the characters, consistent with their condition (i.e., supervisor/perpetrator, subordinate/victim and two other coworkers which were the same across different conditions) and the organization (a hypothetical hotel). After reading the paragraph, participants responded to a multiple-choice quiz question based on that information. (Please see Material and Measures for details about the content.)

During the rest of the week, participants continued receiving e-mails including links to the vignettes once a day about the particular supervisor-subordinate dyad condition they were randomly assigned to. On those days, the participants read one neutral and one abusive supervision vignette in a randomized order and then responded to a multiple-choice quiz question about the content. Conditional on completing the previous day’s requirements, e-mails containing links for the vignettes were sent by 9 a.m. each day, and mini-quizzes were open until 11 p.m. on that night from Monday until Friday. The quiz scores were evaluated cumulatively during data analysis so that falsehood of an answer was not a reason for dropping participants from the study. Participants kept receiving new e-mails as long as they responded to the quizzes.

After finishing all the requirements, i.e., completing the questionnaire and reading the basic information about the characters on the first day, reading one neutral and one abusive supervision vignette on the following four days (eight vignettes in total), and responding to daily quizzes (five in total, one per day), participants received an e-mail containing the link to the final questionnaire. The final questionnaire, involving perceived acceptability and willingness to help scales, was sent on Saturday morning, and had to be filled out by 9 a.m. at the latest on Monday morning, before the start of the new round of data collection. All emails were sent through the university’s online experiment system to ensure anonymity for research participants so that we had only access to the System IDs of participants, which could not be tied to any personal information (see Table 1 for summary of the experimental procedure).

Participants

A total of 441 participants, who were recruited from the student subject pool of an American university, enrolled in the study. Due to the demanding nature of the one-week long procedure, only 207 participants completed all the requirements of the study, with an attrition rate of 53.1%. Those 207 participants received full credits and others received partial credits for their contribution. Three out of five was set as the cutoff point for cumulative quiz scores, and ten of the 207 participants were excluded from analyses due to low scores. Thus, the final sample size was 197, consisting of 42 male and 155 female participants, with a mean age of 20.29 (SD = 3.31). More than half of the participants (53.3%) defined their race as White, followed by Hispanic (18.8%), White non-Hispanic (11.7 %), African American (7.1%), Asian Pacific Islander (5.6%) and other races (3.6%). Other demographics showed that almost half of the participants were freshmen in college (47.2%), followed by sophomores (24.9%), juniors (20.8%), and seniors (7.1%). Finally, the majority of the participants (89.4 %) had full-time or part-time work experience.

Material and Measures

Vignettes. The vignettes were developed for this study and were reviewed by an expert. Pilot testing also did not reveal any problems. As explained previously, vignettes were shared for four consecutive days, each describing a series of day-to-day activities of the previously mentioned characters. Each vignette focused primarily on the relationship between the specific dyad either in the presence or absence of other employees. A total of eight vignettes were developed consisting of four neutral interactions and four interactions involving abusive supervision behaviors in the workplace. The events involved in the vignettes took place in several contexts such as the subordinate’s orientation on her/his first day at work, one-to-one communication between the dyad, monthly meetings attended by other employees, in absence of the subordinate (where the perpetrator was talking negatively about her/him), and in a regular office setting. Vignettes representing abusive supervision behaviors were developed based on Tepper’s (2000) list of abusive supervision behaviors, such as “Tells me my thoughts or feelings are stupid”, “Reminds me of my past mistakes and failures”, and “Tells me I am incompetent.” (see Appendix A).

Demographic questions. The demographics section, except for age, consisted of multiple choice questions: gender (male or female), race (White, Hispanic, White non-Hispanic, African American, Asian Pacific Islander, and other), work experience (full-time, part-time, both or none), and year in college (freshman, sophomore, junior or senior).

Power distance orientation. The scale developed by Earley and Erez (1997), consisting of eight items, was used to measure this construct (α = .68). Items such as “In work-related matters, managers have a right to expect obedience from their subordinates” and “Employees should not express disagreements with their managers” were rated on a scale ranging from one (strongly disagree) to five (strongly agree). A total score based on the mean of item scores was calculated, where higher scores indicated higher power distance orientation.

Daily quizzes. Participants were required to read vignettes on days 1 through 5. After this, they were given a multiple choice quiz question. Each question was designed to measure participants’ comprehension of the material. There were a total of five quizzes. A total quiz score was calculated for each participant based on the number of correct answers. A minimum cut-off point was put in place at a score of three, and then used as a criterion to exclude some of the participants from data analyses.

Perceived acceptability of abusive supervision. Due to lack of any scales that tapped into this construct, this scale was developed during this research design. The first version of the scale consisted of seven items. However, two items were eliminated during factor analysis (revealing a single-factor structure) and item analysis for improving scale reliability. The final scale used for analyses consisted of five items (α = .62). After reversing one item, mean score was calculated, where a higher score indicated a higher perceived acceptability. (See Appendix B).

Helping preferences. The helping preferences scale was developed during the research design, based on previous studies of observers’ reactions to similar situations, such as mobbing or sexual harassment. The first version of the scale consisted of seven items. However, two items were removed during factor analysis and the final scale used for analyses consisted of six items, two measuring direct help (r = .58), three measuring indirect help (α = .63), and one measuring avoidance of helping. All the items were evaluated with regard to the same instruction: “Please indicate your likelihood to engage in the following behaviors, if you were one of [victim’s name] coworkers.”, on a scale ranging from 1 (extremely unlikely) to 7 (extremely likely). (See Appendix C).

Recognition of the situation. A last direct question was included in the questionnaire: “[Perpetrator’s name] is an abusive supervisor.” Participants rated this item on a scale ranging from 1 (strongly disagree) to 5 (strongly agree).

Results

Descriptives

Descriptive statistics and correlations were examined before hypothesis testing (see Table 2). Correlational trends showed that perceived acceptability correlated significantly with all key variables; the coefficients were positive for perceived acceptability and avoidance and negative for direct help, indirect help, and recognition of the situation, in line with the proposed relationships. One-way analysis of variance (ANOVA) was conducted to examine the impact of having work experience (full-time, part-time or both vs. none) on the continuous study variables and results revealed that having work experience had no significant effects on power distance orientation, F(1, 195) = .438, p = .509; perceived acceptability, F(1, 195) = 1.841, p = .176; direct help, F(1, 195) = .011, p = .915; indirect help, F(1, 195) = 1.117, p = .292; avoidance, F(1, 195) = 3.475, p = .064; and recognition of the situation, F(1, 162) = 1.861, p = .174. Further analyses were conducted for examining the impact of gender, specifically due to the imbalanced distribution of male and female participants. Partial correlations showed that the distribution of significant vs. insignificant correlation coefficients did not change when gender was controlled for. Similarly, the values of the coefficients also remained at very similar levels (see Table 3).

Mean values relevant to the proposed relationships were also examined. First, the perceived acceptability scores reported by male and female observers for male vs. female perpetrators were examined (see Table 4). The highest scores were obtained for male observer-male perpetrator condition (M = 1.97, SD = 0.55), whereas the lowest score was obtained for female observer-male perpetrator condition (M = 1.51, SD = 0.46). Comparison of helping scores for male versus female victim conditions (Table 5) showed that for both types of victims mean values were highest for tendency to provide indirect help (M_FemaleVictim = 5.20, SD = 1.04; M_MaleVictim = 5.08, SD = 1.03) and none of the help-related variables were significantly different in the comparison of male versus female victim situations.

Hypothesis Testing

In order to test the hypotheses, several statistical analyses were conducted. First, the percentages of participants based on their responses on recognition of the situation as a case of abusive supervision were compared via chi-square (Hypothesis 1). Second, a path analysis was conducted to test the model suggesting that power distance orientation predicts perceived acceptability of the situation (Hypothesis 2) and perceived acceptability predicts observers’ intention to help (Hypotheses 4). Third, the same model was tested separately for male and female victim conditions to see differences (moderation effects) in helping-related variables via parallel testing (Hypotheses 5 and 6). Finally, a two-way ANOVA was conducted to test main and interaction effects of genders of observer and perpetrator on perceived acceptability of abusive supervision.

According to the first hypothesis, it was predicted that the majority of participants would notice that the supervisor was abusing her/his subordinate. Participants’ rating of the relevant statement (“[Perpetrator’ name] is an abusive supervisor”) was analyzed and results revealed that 75.6% of the participants selected either strongly agree (17.1%) or agree (58.5 %) in their evaluation of this item, whereas 16.5% were indecisive, and 7.9% either disagreed (7.3%) or strongly disagreed (.6%). The percentage of those who accepted the situation as an abusive supervision situation (75.6%) was statistically compared to others’ and the result of chi-square test was significant, χ²(1, N = 197) = 103.023, p < .001. This finding revealed support for the first hypothesis, so that the majority of participants could confidently state that the supervisor was abusing her/his subordinate. Thus, the unusual situation was noticed by those participants, where only 7.9% of participants clearly rejected that the supervisor was abusive. This variable also had a significant negative correlation with perceived acceptability (r = -.21, p < .01), which was expected based on the proposed relationships.

A path model was developed for hypothesis testing and was tested using MPlus software, using true correlations accounting for measurement error. In the original study model (see Figure 1), power distance orientation of observers was expected to predict their perceived acceptability of abusive supervision (Hypothesis 2), and this relationship was expected to be moderated by a match between genders of observer and perpetrator (Hypothesis 3). In the following steps of the model, a higher perceived acceptability was expected to increase the likelihood of avoiding to support the victim (Hypothesis 4a), and to decrease the likelihood of providing indirect support (Hypothesis 4b) and direct support to the victim (Hypothesis 4c). The variable representing observer-perpetrator gender match vs. mismatch was removed from the model due to imbalanced sample sizes in four conditions of observer-perpetrator gender combinations, limiting the statistical soundness of parallel testing (e.g., 21 male observers vs. 84 female observers for male perpetrators; see Table 4), so that Hypothesis 3 could not be tested via path modeling.

The model, including all direct and indirect relationships, was analyzed and the results revealed a good model fit with satisfactory indices based on criteria outlined by Kline (2010) and Hu and Bentler (1999), indicated by an insignificant chi-square value for the covariance matrix, χ²(21, N = 197) = 31.86, p = .061; an RMSEA value lower than .10 (RMSEA = .051, 90% CI [.000-.085]); a CFI value higher than .90 (CFI = .928); a TLI value close to 1 (TLI = .904); and an SRMR value less than .08 (SRMR = .065). Given the good model fit indices, coefficients for specific relationships were examined (see Figure 2 for coefficients). According to the results of path analysis, power distance orientation of observers predicts their perceived acceptability of abusive supervision (β =.19, p = .005), so that higher power distance leads to higher acceptability, consistent with Hypothesis 2. Moreover, higher perceived acceptability leads to a higher tendency to avoid intervention (β = .65, p = .001), and a lower tendency to provide direct support (β = -.59, p = .000), and indirect support (β = -.71, p = .000) to the victim. Thus, Hypotheses 2, 4a, 4b, and 4c were supported. The direct and indirect (mediated) relationships between power distance orientation and helping variables were examined via the procedure recommended by Baron and Kenny (1986). Results revealed that power distance orientation had no significant direct effect on direct help, indirect help, and avoidance, so that all the effects were mediated via perceived acceptability.

The path model developed for testing Hypotheses 2 and 4 was later tested separately for male and female victim conditions to test relevant hypotheses stating moderation by gender of the victim. Specifically, the effect of perceived acceptability on avoidance (Hypothesis 5) and on indirect help (Hypothesis 6b) were predicted to be stronger for male victims, whereas its effect on direct help was predicted to be stronger for female victims (Hypothesis 6a). The first model included only female victim conditions (N = 106) and model fit was satisfactory, based on the following indices: χ²(8, N = 101) = 14.242, p = .076; RMSEA = .086, 90 % CI [.000, .157]; CFI = .929; TLI = .866; and SRMR = .079. The second model was tested for male victim conditions only (N = 91) and the model fit was satisfactory, based on the the following indices: χ²(8, N = 91) = 9.413, p = .309; RMSEA = .044, 90 % CI [.000, .135]; CFI = .978; TLI = .959; and SRMR = .064. In both models, the relationship between power distance orientation and perceived acceptability was significant, (β_FemaleVictim =.23, p = .026; β_MaleVictim = .17, p = .047), consistent with the first model developed for hypothesis testing. Results for the relationships between perceived acceptability and help-related variables showed some variations, so that in female victim conditions higher perceived acceptability increased avoidance to help the victim (β = .77, p = .003), and decreased tendencies for providing indirect help to the victim (β = -.78, p = .000), and proving direct help to the victim (β = -.82, p = .000). However, in male victim conditions, only the relationship between perceived acceptability and tendency to provide indirect help was significant (β = -.63, p = .006), where perceived acceptability decreased willingness to provide indirect help (Figure 3 and 4).

Comparison of avoidance scores for male (M = 3.66, SD = 1.38) vs. female victim (M = 3.59, SD = 1.41) conditions revealed insignificant results, t(168) = -0.288, p = .773. Thus, Hypothesis 5 was not supported, since this effect was predicted to be stronger for male victims. Results of parallel testing suggested that for female victims, perceived acceptability was a significant factor in determining observers’ willingness to support them (either directly or indirectly) or to avoid supporting them. For male victims, the effect of perceived acceptability on avoidance was insignificant (β = .47, p = .139), showing that it was only a factor explaining avoidance behaviors in female victim cases. Hypothesis 6a was supported since the negative effect of perceived acceptability on willingness to provide direct help was strong and significant for female victims (β = -.82, p = .000), whereas it was weak and insignificant for male victims (β = -.21, p = .362). Contrary to Hypothesis 6b, the effect of perceived acceptability on indirect help was stronger for female victims (β = -.78, p = .000) in comparison to male victims (β = -.63, p = .006).

Hypothesis 3 stated that the relationship between power distance orientation and perceived acceptability of abusive supervision is stronger when observer and perpetrator are of the same gender. As explained before, this moderation hypothesis could not be tested via parallel testing in path modeling due to the imbalanced sample sizes. The direct and interaction effects of observers’ gender and victims’ gender were examined separately via two-way ANOVA for exploratory purposes. In this analysis, the dependent variable was perceived acceptability of abusive supervision and the independent variables were perpetrator’s gender, observer’s gender, and their interaction. The results showed that perceived acceptability of abusive supervision was predicted by observers’ gender, F(1, 193) = 8.434, p = .004, and the interaction between perpetrator’s and observer’s gender, F(1, 193) = 6.697, p = .010 (see Table 4 for specific mean values). The interaction was further examined via a figure, where the impact of gender match on perceived acceptability was clearly illustrated for male observers, so that male abusive supervision by male perpetrators was perceived to be relatively acceptable especially by male observers, whereas perception of abusive behaviors by female perpetrators was relatively stable among male and female observers (see Figure 5).

Exploratory Analyses

The hypothesized relationships in the study focused on genders of perpetrator and observer in the analysis of perception of abusive behaviors, in relation to attitudes toward the power figure and the gender of the victim in the analysis of willingness to help the victim, in relation to attitudes toward male vs. female victims in need of help. Exploratory regression analyses were conducted with the purpose of analyzing the effects of the interaction of victims’ and observers’ genders on avoidance, direct help, and indirect help as three separate criterion variables. Predictors for all regression analyses were defined as perceived acceptability and four dummy coded variables for different observer-victim dyads: (i) female observer-female victim, (ii) male observer-male victim, (iii) female observer-male victim, and (iv) male observer-female victim. For these dummy-coded variables, all conditions other than the target condition were coded as zero, where the target variable was coded as 1.

Multiple regression was calculated to predict avoidance to provide help based on the five variables (perceived acceptability and four dummy variables). A significant regression equation was found, F(4, 165) = 8.117, p = .002; however among all the variables, only perceived acceptability was found as a significant predictor (β = .306, p = .000), so that perceived acceptability increased avoidance to help. The regression equation for direct help was also significant, F(4, 164) = 6.299, p = .000, and significant predictors were perceived acceptability (β = -.351, p = .000) and male observer-female victim condition (β = .160, p = .042), so that higher perceived acceptability decreased willingness to help the victim directly, and male observers of abusive supervision situations where the victim was female were more likely to provide direct help in comparison to other dyadic conditions. Finally, the regression equation for indirect help was also significant, F(4, 192) = 7.462, p = .000, and significant predictors were perceived acceptability (β = -.321, p = .000) and male observer-male victim condition (β = -.150, p = .037), so that a higher perceived acceptability decreased willingness to help the victim directly and male observers witnessing male victims being abused were less likely to provide indirect help.

Discussion

Results of the study shed light on the abusive supervision phenomenon based on observers’ perception and potential intervention with regard to the stages of the Bystander Intervention Framework (Latané & Darley, 1970), consistent with the purpose of the study. Specific findings will be discussed in line with the five stages of the framework, which were explained previously. At the first stage, observers were expected to notice the problematic situation and the majority of participants of the present study labeled the perpetrator as an “abusive supervisor”, consistent with this expectation. Thus, they acknowledged the presence of a situation where a subordinate was being abused by his/her supervisor.

The second stage of the framework is concerned with the interpretation of the situation. Based on the results of the study, individual-level power distance orientation (Earley & Erez, 1997; Maznevski et al., 2002) of observers is one of the main determinants of perceived acceptability of the abusive supervision situation, consistent with studies focusing on perception of abusive supervision by victims (Lian et al., 2012; Wang et al., 2012). In sum, observers who were expecting a gap between the levels of power held by superiors vs. subordinates and who respected the authority of power figures were less likely to question the abusive supervision situation and more likely to perceive it to be acceptable.

As a second key variable in this stage, perpetrator’s and observer’s gender match played a role in the interpretation of the situation. As explained before, the hypothesized moderating effect of this match on power distance orientation could not be tested due to sample size issues. Nevertheless, the direct effect of gender match on perceived acceptability was clear, specifically for male observers, so that their tolerance was much higher for male perpetrators in comparison to female perpetrators. This finding can be explained via the Social Identity Theory (Tajfel, 1974), given that male observers were more tolerant toward abusive behaviors of their in-group members (male perpetrators) in comparison to out-group members (female perpetrators). The lack of a similar effect for female observers can be explained by findings of studies on women in management, which show that female employees are not more tolerant, and may be even harsher, in their judgments of female managers (Sheppard & Aquino, 2013), lacking a spirit of “sisterhood” (Mavin, 2006).

In the third stage of the framework, observers decide if they should take responsibility or not. In line with the relevant hypothesis, perceived acceptability of abusive supervision appeared to be a determinant of taking action so that participants who perceived the situation to be more acceptable were more likely to avoid intervention, while the ones who perceived the situation to be less acceptable were more likely to provide help to victims, direct or indirectly. Contrary to the hypothesis, male victims were not less likely to receive help, in comparison to female victims, so that observers’ levels of likelihood for avoiding intervention were in similar levels for male vs. female victims. However, the effect of perceived acceptability on avoidance was not significant for male victim conditions, indicating that there might be other factors impacting observers’ willingness to help male victims. For example, the degree of benevolent sexism (Glick & Fiske, 1997) may play a role, given that high benevolent sexism tendencies induce protection of women rather than men.

In the fourth stage, observers who are willing to help are expected to decide on how to support the victim. In the present study, a distinction between direct vs. indirect helping behaviors was made and the results revealed that a victim’s gender impacted the effect of perceived acceptability on tendencies toward supporting the victim directly or indirectly. The effect was strong and significant for female victims, whereas it was weak and insignificant for male victims. Thus, whether one witnesses a male or female coworker being abused makes a difference in that female victims are likely to receive direct help from observers, whereas male victims are not likely to receive direct help, regardless of the level of perceived acceptability. This finding was consistent with hypothesized relationships based on premises of the concept of benevolent sexism (Glick & Fiske, 1997). Specifically, female victims might be more likely to be perceived as weak by observers and helping them directly probably does not seem so much of a threat to social order, whereas male victims deserve indirect help only, given that they can/should protect themselves, and providing direct help may be seen as a threat to their masculine image. The impact of perceived acceptability on willingness to help the victim was also stronger for female victims, in comparison to male victims, but it did not lose its significance in male victim conditions, contrary to the case of direct help. Based on the results of exploratory analyses, among the four conditions representing different observer-victim dyads, in the analysis of direct help only male observers of female victim conditions had the tendency to provide help, whereas in the analysis of indirect help male observers of male victim conditions had a lower tendency to provide help. Thus, an observer’s gender may play a role here as well, though these comparisons should be interpreted cautiously due to the imbalanced sample sizes in the four different dyadic conditions.

In the final stage of the Bystander Intervention Framework, tendencies toward helping and avoiding intervention are expected to translate into actual behaviors, consistent with the Theory of Planned Behavior (Ajzen, 1991). Due to the scope and design of this study, this final stage could not be tested with this sample.

Theoretical Implications

The previous studies on perception of abusive supervision by third-party observers examined several components of an observer’s role in these situations, focusing mainly on observers’ characteristics determining their perceptions (e.g., Lian et al., 2012; Wang et al., 2015; Wang et al., 2012), without any underlying comprehensive theory. However, adopting a well-established theory may help us understand the decision making process better with a comprehensive approach. The Bystander Intervention Framework (Latané & Darley, 1970) has been applied to several contexts before, such as bullying in schools (Polanin et al., 2012) and sexual violence in different contexts (McMahon & Banyard, 2012). This study represents the first attempt for applying this framework to abusive supervision situations, and the results suggest that using this approach may enrich our comprehension of third-party observers’ experiences in abusive supervision cases, e.g., based on gender-based perceptions and personal orientations such as power distance. Utilization of this framework as a basis for examining an observer’s role and involvement in abusive supervision may guide further research on this topic, via experimental or field studies.

Methodological Implications

A major strength of this study was research methodology. The experimental design used in this study is not a frequently used method and it serves as an alternative to cross-sectional studies where participants are expected to read one vignette and respond to several questions, based on their judgments immediately after reading the vignette. This study, due to its extensive nature, gave participants the chance to form impressions over a one-week period, with regard to several incidents including both abusive supervision and neutral incidents. This method also provides the opportunity to control for several factors, such as the number of bystanders (Latané & Darley, 1970), which may impact perceptions of bystanders. Despite the strong design characteristics, the methodology was under risk due to low scale reliabilities, indicating that there is need for good measures for examining observer perceptions. Thus, as another implication, the scales used in this study may guide further efforts for developing sound scales for measuring acceptability of abusive supervision and willingness to help the victims.

Practical Implications

Based on results of the present study, three key lessons can be drawn for practice. First, organizations should be careful about power distance orientation at individual, organizational, and cultural levels, since high levels of this orientation may lead to higher acceptability of abusive supervision by observers, as well as victims (Lian et al., 2012; Wang et al., 2012). Second, male victims of abusive supervision may need additional organizational support, given that they were less likely to receive direct help from observers. Third, organizations should be particularly careful about potential male perpetrators, since especially male observers were more tolerant towards their abusive behaviors.

Building on these issues, organizations can use two key mechanisms in their effort to minimize abusive supervision in work context. First, they should be careful about developing sustainable organizational cultures which promote and preserve ethical work environments. Organizational climates and cultures supporting high power distance may be a risk factor against the strive for ethical work places where behaviors such as abusive supervision are sincerely discouraged. Similarly, given that gender plays a role in the perception of abusive supervision, male dominant work places and organizational cultures may cause a risk to ethical climate and be more open to the negative impacts of abusive supervision. Thus, acknowledging culture at national level, which may be high in power distance and gender inequality, organizations should put effort for better work places by actively promoting ethical behavior and diversity. In addition to culture and climate, formal policies of organizations should be also supporting the ideals of ethical behavior and encourage observers, as well as victims, to speak out. Presence and application of such clear policies would be discouraging for perpetrators as well. Policies might be particularly helpful for male victims, who might be more likely to receive indirect help via organizational routes, where policies play a critical role. This effort may also have positive implications for employees with different sexual orientations, going above binomial assumptions of gender groups.

Limitations

The study had five main limitations. First, being vignette-based may be seen as a weakness, since it may limit generalizability of results to real life settings. However, as mentioned before, it provided the opportunity to control for several variables and focus on the effects of gender. Since abusive supervision cannot be manipulated in a real context, vignette-based research design is an accepted model of research in this area (e.g., Farh & Chen, 2014; Park et al., 2018). Second, the sample consisted of students due to the demanding experimental nature of the study; this may be considered a risk against generalizability of results to employees in real work settings. However, 89.4% of the sample had full- or part-time work experience, thus they were expected to have developed a notion of work contexts and norms in work settings. As a third limitation, as explained before, the low number of male respondents in the student subject pool impacted the analyses. Due to the demanding nature of the experimental design, recruiting the participants and retaining them for one week was a major challenge and the ratio of male participants could not be improved. Nevertheless, significant results, despite this limitation and imbalance among different conditions, may signal important relationships, which necessitate further research.

The fourth limitation was related to the order of presentation of the information and questions in line with the stages of the Bystander Intervention Framework (Latané & Darley, 1970). The most important issue here may be that the critical direct question about observers’ perception of the situation was only asked on the last day of data collection. Specifically, participants were asked to rate the statement “Jane/John is an abusive supervisor” to understand if they recognized the situation, consistent with the first step of the framework. Asking this question on the first day (rather than on the last day) would be a serious problem, since this direct question would prime participants about the “abuse” aspect of upcoming vignettes. As the fifth limitation, reliability scores of the scales developed for this study were not very strong with values around .60, which may therefore cause a risk in the measurement. However, it is also suggested that .60 can be an acceptable value for the alpha coefficient (Loewenthal, 2004). Nevertheless, these scales should be improved for further research on this topic.

Further Research

There is potential for further research on the perception and role of observers of abusive supervision, both conceptually and methodologically, with regard to the findings of this study. Conceptually, there may be many different factors impacting perceived acceptability of the situation and willingness to support the victims, in addition to study variables such as power distance orientation and genders of observer, victim, and perpetrator. For example, research shows that negative affectivity of subordinates plays a role in their perception of abusive supervision (Tepper et al., 2006), leading to a more pessimistic view where they find it less acceptable and are more influenced by its negative effects, whereas positive affectivity increases proactive behavior against abusive supervision (Xu et al., 2019). Similarly, victims with high levels of agreeableness, extraversion (Wang et al., 2015), and aggression (Brees et al., 2014) are more likely to react to abusive supervision (Wang et al., 2015). These personality characteristics may impact observers as well.

Including benevolent sexism (Glick & Fiske, 1997) as a variable, which may impact perceptions and actions of observers, may explain observers’ gender-based attitudes toward victims. Personal relationships with the victim should be also considered since observers are more likely to provide help if they feel close with the victim (Coyne et al., 2019). Similarly, observers’ relationship with the perpetrator may also play a role (Pan & Lin, 2016). Investigation of the ethical climate (Cullen et al., 1989) prevalent in organizations may be promising as a field study, given that it may impact observers’ perceptions of abusive supervision and also their willingness to take action. Moreover, the analysis of implications of power distance orientation at national level via a cross-cultural study may widen our theoretical perspective, since behaviors such as bullying may be perceived to be more tolerable in some cultures, such as Italy (Giorgi et al., 2015). Finally, the effect of gender similarity with the perpetrator for victims vs. observers can be compared in future studies; a recent study showed that victims of abusive supervision experience higher levels of psychological distress when a perpetrator’s gender matches their gender, since they feel higher levels of disappointment (Park et al., 2018).