Abusive supervision and turnover intention among public servants: the roles of psychological distress and person-organization fit

ABSTRACT

This study integrates social cognitive and cognitive dissonance theories to examine the influence of abusive supervision on public servants. By employing Hayes’ Process macro and investigating a cohort of 468 US public servants, we found that abusive supervision is positively related to turnover intention via psychological distress. Notably, the adverse impact of abusive supervision is more pronounced among subordinates with a strong person-organization fit. Given the prevailing retention challenges in the public sector, our research offers novel insights into retaining well-matched personnel through the cultivation of positive workplace social connections.

KEYWORDS:

• Abusive supervision
• psychological distress
• turnover intention
• US public sector

Productive and motivated public servants are crucial for achieving public service quality and public sector missions (Eriksson and Andersson 2023). However, retaining high-performing public servants poses a formidable challenge, notably during the recent ‘Great Resignation’ in the US (Bao and Zhong 2021; Corin, Berntson, and Härenstam 2016; Jansen 2022; Kaufmann, Borry, and DeHart-Davis 2023). The elevated turnover rate among public servants carries significant implications, affecting the quality of service delivery and the performance of public sector organizations (Moon 2017). The financial impact is accentuated in the public sector due to its more intricate and resource-intensive recruitment processes (Jansen 2022). Therefore, a comprehensive investigation into the departure of productive public servants is imperative to inform management strategies aimed at nurturing a high-quality public sector workforce (Wynen et al. 2019).

Public Administration (PA) research has identified various internal and external factors influencing public servants’ turnover intention. Studies have established links between turnover intention and individual factors such as public service motivation (PSM) and job satisfaction (e.g. Bright 2007, 2008, 2021; Gould-Williams, Mostafa, and Bottomley 2015; Liu, Tang, and Yang 2015; Quratulain and Khan 2015; Shim, Park, and Eom 2017). Other PA research (Boon, Wynen, and Kleizen 2021; Campbell and Im 2016; Jung, Chan, and Hsieh 2017; Kaufmann, Borry, and DeHart-Davis 2023) has explored the impacts of organizational changes, psychological climates, and rules on turnover intention. Despite the extensive scope of turnover intention research in PA, there is a notable gap highlighted by Hur and Abner (2023) regarding the need to delve into work environment conditions, an area that remains insufficiently studied. A critical area for advancing research is the role of leadership as a contextual factor within the workplace that influences turnover intention (Hur and Abner 2023). The prevailing focus on transformational leadership has overshadowed other leadership styles within the PA literature (Backhaus and Vogel 2022). PA scholars (Backhaus and Vogel 2022; Hassan 2019) have advocated for an increased exploration of harmful leadership behaviours in the public sector due to their severe repercussions for public servants. This study responds to this call by addressing the research question: How do turnover intention among public servants occur from an unethical leadership perspective? Specifically, our study delves into the mechanisms of the relationship between abusive supervision and turnover intention, thus contributing to the paucity of PA research on detrimental leadership behaviours and their impact on turnover intention.

The study’s theoretical framework is grounded in Social Cognitive Theory (SCT) (Bandura 1986, 2001). SCT asserts that the specifics of a work environment, including interpersonal relationships and leadership behaviours, influence individuals’ sense-making about their work and environment, thereby shaping their behaviours (Bandura 1986, 2001). Notably, leadership behaviours serve as crucial social cues for public servants, internalizing work’s importance, and self-worth’s value, thus moulding attitudes and behaviours (Vogel and Masal 2015). As abusive leadership behaviours transgress moral standards, public servants may perceive uncertainty and ambiguity regarding their jobs, personal growth, and self-concept (Schaubroeck, Peng, and Hannah 2016). The resulting psychological distress, encompassing anxiety, fear, and depression, arises from the interpretation of the work environment as stress-inducing and hostile (Fischer et al. 2021; Martinko et al. 2013; Tepper 2000). Consequently, distress from exposure to abusive supervision may cultivate intention to leave the organization (Rice, Taylor, and Forrester 2020). In summary, building upon SCT, we posit positive associations among abusive supervision, psychological distress, and turnover intention.

While SCT provides a fundamental basis for understanding public servants’ turnover intention under abusive supervision, it may not fully elucidate how they uniquely develop turnover intention. There are disparities in public servants’ reactions to abusive supervision behaviours (see Fischer et al. 2021; Mackey et al. 2017; Martinko et al. 2013; Tepper, Simon, and Park 2017; Zhang and Liao 2015). As the conditioned process through which abusive supervision relates to the turnover intention of public servants has received little attention in the PA research, we incorporate the lens of SCT into the perspective of person-organization (P-O) fit through Cognitive Dissonance Theory (CDT, Festinger 1957) to examine how P-O fit can attenuate or accentuate the consequences of abusive supervision.

CDT posits that cognitive dissonance, or an unpleasant condition, occurs when individuals experience contradictory cognitions or disparities between their beliefs, values, attitudes, and behaviours and those of the organization (Festinger 1957). In this instance, they will seek to minimize the magnitude of dissonance, such as changing the interpretation of work events or adding new attitudes, behaviours, or beliefs (Festinger 1957). The applicability of CDT to explain the moderating effect of P-O fit on the consequences of abusive supervision is relevant because leaders create the workplace context, and their behaviours are seen as the reflection of organizational values and norms. Highly P-O-fitted employees have value congruence with organizations; thus, they become more vulnerable to unmet-expectation situations (Kristof-Brown, Zimmerman, and Johnson 2005; Mostafa et al. 2022; Ruiz-Palomino and Martínez-Cañas 2014; Verquer, Beehr, and Wagner 2003). In line with CDT tenets, public servants with high P-O fit could interpret the abusive behaviours of their supervisors as a meaningful discrepancy between their values and those of the organization. They are more likely to look for an ‘ideal workplace’ to satisfy their needs for value congruence or have higher levels of turnover intention (Ambrose, Arnaud, and Schminke 2008; Thorne 2010).

We choose to study P-O fit as a moderator since it is a pertinent topic in PA research, associated with public value embracement and the motivation to fulfil social responsibilities to serve the public good (Karolidis and Vouzas 2019; Perry and Hondeghem 2008; Wright and Pandey 2008). Additionally, while the lens of P-O fit as a moderator is crucial and impactful, there is a lack of research on it, leading to a missing piece of the work environment-employee puzzle. Much of the previous research in PA has predominantly explored P-O fit as a mediator within the association between public service motivation and work-related outcomes (Bright 2021; Gould-Williams, Mostafa, and Bottomley 2015; Kim 2012; Wright and Pandey 2008). However, emerging literature underscores the pivotal role of value congruence between the organization and its employees in elucidating the impact of workplace conditions (Kristof-Brown, Treviño, and Harrison 2005; Mostafa et al. 2022; Ruiz-Palomino and Martínez-Cañas 2014; Verquer, Beehr, and Wagner 2003). Given this research gap, our study aims to contribute to a more comprehensive understanding of P-O fit’s moderating effect, highlighting its importance in bridging this divide. Lastly, the persistent issue of turnover in public sector organizations, along with its implications for public service performance, underscores the urgency of comprehending the moderating effect of P-O fit. Such understanding can equip managers with valuable insights into how high-performance public servants, or those with a strong P-O fit, may respond to dysfunctional or stressful work environments.

This study contributes to the existing PA literature in three major aspects. First, it extends the limited and underdeveloped domain of PA research concerning harmful leadership (Backhaus and Vogel 2022; Hassan 2019), while concurrently addressing the significant research gap in associating such harmful leadership with turnover intention (Hur and Abner 2023). We provide insights into the repercussions of abusive supervision on mental well-being (i.e. psychological distress), subsequently driving turnover intention of public servants, within the context of prevailing research gaps (Backhaus and Vogel 2022; Hassan 2019; Hur and Abner 2023).

Second, we enhance comprehension of the abusive supervision – psychological distress – turnover intention nexus within the public sector by investigating the moderating influence of P-O fit. Prior PA research has suggested that public servants closely aligned with their organizational values (high P-O fit public servants) are more inclined to remain (Bright 2008; Gould-Williams, Mostafa, and Bottomley 2015; Jin, McDonald, and Park 2018; Kim 2012; Moynihan and Pandey 2008). Our study contributes to this meaningful line of inquiry by investigating the moderating role of P-O fit in explaining the abusive supervision – turnover intention linkage. Rooted in the perspective of CDT, our findings indicate that public servants with a strong P-O fit exhibit heightened sensitivity to supervisory abuse, resulting in elevated psychological distress and a corresponding rise in turnover intention. These findings shed light on the reasons highly P-O-fitted public servants might choose to depart, affording management insights for retaining productive personnel.

Finally, our study advances emerging literature using SCT in the public sector. SCT helps explain how an environmental factor (abusive supervision) influences a public servants’ behaviour (turnover intention) via an intrapersonal process (psychological distress). However, SCT falls short of explaining how individuals may react differently to abusive supervision. Such an approach may prevent an insightful understanding of the consequences of destructive leadership phenomena in the public sector. Our study is novel in that we integrate SCT and CDT to study how the interplay between individuals and workplace environment elements determines public servants’ functioning. The findings offer a more comprehensive understanding of public servants’ behaviours and advance the use of SCT in the public sector context.

Literature review and hypothesis development

Theoretical background

SCT (Bandura 1986, 2001) serves as the primary theoretical framework that explains how individuals may adopt specific behavioural patterns. SCT posits that human functioning is influenced by the interactive dynamics among personal, environmental, and behavioural factors, which constantly reciprocate and shape one another. In this framework, employees process information from their work environment and interpersonal relationships to acquire and internalize normative standards and expectations for appropriate attitudes and behaviours (Bandura 1986, 2001; Schunk and DiBenedetto 2020). Alongside this social learning process, employees also envision the outcomes of their intended reactions to workplace situations to ensure consistent behaviour. A core tenet of SCT is the cognitive process in which employees activate self-regulation and reflection to align their attitudes and behaviours with the demands of their work environment (Bandura 1986, 2001; Schunk and DiBenedetto 2020).

Studies conducted in various contexts, including education (Rubenstein et al. 2018), general management (Otaye-Ebede, Shaffakat, and Foster 2020), and public administration (Nguyen et al. 2022; Park and Hassan 2018), have applied SCT to investigate how employees respond to leadership behaviours and workplace experiences. Grounded in SCT, these studies argue that employees’ behaviours result from the influences of leadership and workplace situations (as environmental factors) mediated through intrapersonal psychological cognition processes (Bandura 2001). In alignment with SCT’s core tenets, we propose that abusive supervision, perceived as a threat by public servants, functions as an environmental catalyst that triggers distress (an intrapersonal affective state) (Mawritz et al. 2012; Tepper 2000; Tepper, Simon, and Park 2017; Zhang et al. 2014). These uncomfortable and insecure feelings stemming from abusive supervisors’ treatment subsequently contribute to heightened turnover intention (approximating voluntary turnover, Sun and Wang 2017), allowing public servants to mitigate the adverse effects of abusive supervision (Harvey et al. 2007; Tepper, Simon, and Park 2017).

Although SCT could help explain the psychological mechanism wherein abusive supervision influences public servants, SCT does not account for individual differences in engaging in specific behaviours to react towards the abusive behaviours of supervisors. The adverse impacts of abusive supervisors on public servants can be exacerbated or lessened by other factors (Harvey et al. 2007; Jiang, Wang, and Lin 2016; Tepper 2000). The CDT lens (Festinger 1957) could compensate for this limitation of SCT to understand in which situations abusive supervision is most strongly associated with psychological distress and, in turn, turnover intention. According to CDT, individuals are prone to react negatively and unfavourably to workplace situations when they perceive conflicting cognitions (Festinger 1957; Hinojosa et al. 2017). From the perspective of high P-O fit, public servants actively seek alignment between their values, norms, and beliefs with those of the organization. This alignment serves as a compelling factor, attracting and motivating them to commit to and remain with the organization (Gould-Williams, Mostafa, and Bottomley 2015; Jin, McDonald, and Park 2018). However, when a perceived inconsistency arises, particularly due to abusive supervisors, public servants with high P-O fit may experience heightened discomfort and dissatisfaction compared to their counterparts with lower P-O fit, who might display greater tolerance for abusive supervision. Abusive supervisors create an unfair, hostile, and aggressive work environment, amplifying the perceived cognitive dissonance and strengthening the intention to leave as a means to reduce this dissonance. Within the framework of CDT, we argue that P-O fit serves as a pivotal boundary condition that distinguishes the effects of abusive supervision on public servants. The proposed model of this study is depicted in Figure 1.

Figure 1. Proposed model.

Abusive supervision, psychological distress, and turnover intention

Abusive supervision has been well-documented to create detrimental work outcomes as it increases an ‘unbalanced or uncoordinated’ relationship between a supervisor and subordinates (Moin et al. 2022; Saleem et al. 2021; Tepper 2000; Tepper, Simon, and Park 2017). Based on SCT principles, public servants process social information of ridiculed and unjust treatment of abusive supervision as an interpersonal stressor (Zhou et al. 2020) as well as a threat to self-worth and personal growth. This is because the authority of supervisors can influence crucial aspects of their career opportunities (Brown, Treviño, and Harrison 2005; Jiang et al. 2021, Peltokorpi 2019; Tepper, Simon, and Park 2017). Public servants may, therefore, reconsider their association with the organization because abusive supervisory behaviours symbolize organizational values, norms, and goals. The PA literature highlights the critical role of leadership in promoting public servants’ commitment and engagement in delivering public values because leadership behaviours facilitate public servants’ internalization of such values into their performance (Tavares, Sobral, and Wright 2021). Therefore, when public leadership is not role-modelling of ideal principles of the public sector, public servants could feel detached from the organization and intend to leave (Tavares, Sobral, and Wright 2021). Public servants with abusive supervision may perceive a breach of psychological contract with the organization (Martinko, Sikora, and Harvey 2012; Saleem et al. 2021), which influences turnover intention (Rubenstein et al. 2018), as a flight reaction to abusive supervision to minimize the threat from their abusers’ power and authority (Tepper et al. 2009; Zhang et al. 2022).

Hypothesis 1.

Abusive supervision is positively related to turnover intention.

According to SCT, public servants’ attitudes and behaviours are the product of the cognitive process of social information and reflection on workplace situations (Bandura 2001). Critically, the quality of interactions with supervisors is pertinent to human functioning (Bandura 2001; Frone 2000). The general management literature implies that leaders control the resources granted to their subordinates and frame their subordinates’ experience with work and navigation of the workplace environment as they are part of the social environment (Lysova et al. 2019). Therefore, subordinates appreciate a high-quality relationship with their supervisors to maintain positive cognitions, such as motivation, enthusiasm, energy, and satisfaction (Li et al. 2016; Lysova et al. 2019; Yagil et al. 2011). Abusive supervision as the social cue of power misuse threatens public servants’ perceptions of safety and security (Hershcovis and Barling 2010). Furthermore, abusive supervision is associated with perceptions of organizational injustice as public servants are treated unfairly and harmfully (Hoobler and Hu 2013; Tepper 2000; Tepper, Simon, and Park 2017; Zhang et al. 2019). In such unjust and stressful workplaces, public servants will arouse psychological distress as a form of anxiety, fear of punishment, and depression (Liang et al. 2018; Tepper 2000) as they need to deploy personal and psychological resources to navigate these negative workplace dynamics, ultimately leading to strains (Harms et al. 2017; Restubog et al. 2011; Tepper et al. 2007).

Hypothesis 2:

Abusive supervision is positively related to psychological distress

The positive association between cognitively stressful experiences and turnover intention has been noticeable. Evidence shows that public servants tend to leave their organizations due to emotional exhaustion or depersonalization (Kim 2005, 2015). High job demands and work stressors are the reasons for turnover intention among public servants because they develop negative cognitions in interpreting workplace environments as stressful and demanding (Shim, Park, and Eom 2017). Another example of this association is evident in the study of Gamassou (2015), which argued for the need to operate supportive mechanisms in public sector organizations to reduce the experience of stress as one of the critical drivers of turnover intention. Examined by Saleem et al. (2021) with a sample of police officers in Pakistan, psychological distress is positively related to turnover intention within a circumstance with abusive supervision. One would expect that public servants with abusive supervisors may have insufficient time and energy to cope actively with abusive behaviours from the powerful authority that trigger psychological distress (Martinko et al. 2013; Tepper et al. 2009; Zhang et al. 2022). Consequently, having an intention to leave could be an optimal coping strategy for those in such a stressful and insecure situation to protect themselves (Saleem et al. 2021). As such, we argue that public servants with abusive supervisors who experience psychological distress will develop turnover intention.

Hypothesis 3.

Psychological distress is positively related to turnover intention.

The moderation effect of P-O fit

P-O fit broadly refers to the compatibility between individuals and organizations (Kristof 1996, 49:3). Accordingly, P-O fit occurs when (1) both individuals and organizations share congruent values and goals (supplementary fit) and/or (2) either individuals or organizations meet the expectations and satisfy the needs of the other party (complementary fit) (Kristof-Brown, Schneider, and Su 2023). P-O fit is a pertinent topic in the PA literature because it reflects unique aspects of public service and fulfilling social responsibilities (Gould-Williams, Mostafa, and Bottomley 2015; Moynihan and Pandey 2008; Wright and Pandey 2008). Although P-O fit and PSM can produce equivalent outcomes for public sector organizations, unlike PSM, which only focuses on public servants’ predisposition (Perry and Wise 1990), P-O fit emphasizes the alignment of values between public servants and public sector organizations (Cable and DeRue 2002). P-O fit is the driver motivating public servants’ cooperation and collaboration to achieve a broader strategic mission of the organization (Kristof-Brown, Schneider, and Su 2023).

PA literature predominantly focuses on investigating the favourable outcomes associated with P-O fit. The favourable evidence encompasses a range of positive impacts, such as job satisfaction, work performance, organizational citizen behaviour, organizational commitment, and meaningful work (Bright 2008; Gould-Williams, Mostafa, and Bottomley 2015; Hue, Thai, and Tran 2022; Jin, McDonald, and Park 2018; Kim 2012; Mostafa and Gould-Williams 2014; Teo et al. 2016; Thuy and Phinaitrup 2021; Wang and Brower 2019). A high P-O fit fosters alignment with public values and intensifies public servants’ motivation to fulfil the public sector organizational goals of serving the public good (Karolidis and Vouzas 2019; Perry and Hondeghem 2008; Wright and Pandey 2008). Therefore, P-O fit is positively related to job choices, motivation to stay, and work engagement (Bright 2007; Erdogan and Bauer 2005; Gould-Williams, Mostafa, and Bottomley 2015; Jin, McDonald, and Park 2018; Moynihan and Pandey 2008) and negatively related to emotional exhaustion, stress, and turnover intention (Gould-Williams, Mostafa, and Bottomley 2015; Liu, Liu, and Hu 2010; Mostafa 2016).

Although the extant PA research supports the notion that public servants with high P-O fit can overcome stressful situations and tend to be more emotionally attached to the organization, past studies have overlooked the possible impact of employees’ (in)congruent attitudes regarding various workplace circumstances (Chatman 1991; O’Reilly, Chatman, and Caldwell 1991). Based on the aspect of fit, public servants with high P-O fit believe that their values and norms are compatible with those of the organization when these are regulated, endorsed, and followed by other organizational members (Chatman 1991; O’Reilly, Chatman, and Caldwell 1991). Also, these public servants develop trust in the organization in maintaining the standards of what is (un)acceptable and what is the right thing to do (Edwards and Cable 2009). As public servants with high P-O fit associate the values of the organization with the attitudes and behaviours of their working peers and supervisors (Gould-Williams, Mostafa, and Bottomley 2015), they are more susceptible to the potential breach in value compatibility (Kristof-Brown, Treviño, and Harrison 2005; Mostafa et al. 2022; Ruiz-Palomino and Martínez-Cañas 2014; Verquer, Beehr, and Wagner 2003). This notion can be explained by the tenets of CDT (Festinger 1957).

Public servants with a high P-O fit are more willing to spend effort to develop their careers within organizations having similar values and characteristics (Gould-Williams, Mostafa, and Bottomley 2015; Thuy and Phinaitrup 2021). Public sector organizations could espouse values representing prosocial and humanistic exemplars to attract individuals with similar values and goals and guide their appropriate and normative behaviours (Van der Wal, De Graaf, and Lasthuizen 2008). Public servants tend to observe the behaviours of supervisors to rationalize their choice to work at the organizations because supervisors are the organizational representatives (Martinko, Sikora, and Harvey 2012; Saleem et al. 2021). Therefore, it is likely that public servants with high P-O fit look for supervisors to recognize their potential, understand their issues, and provide them with resources when necessary. Doing so satisfies their needs and preferences for their identification with the organization and career opportunities (Brown, Treviño, and Harrison 2005; Jiang et al. 2021; Tepper, Simon, and Park 2017).

Based on the CDT, we argue that public servants with high P-O fit are perceptive to the potential of inconsistency or meaningful discrepancy in social information about supervisory behaviours and organizational values and characteristics (Kristof-Brown, Treviño, and Harrison 2005; Mostafa et al. 2022; Ruiz-Palomino and Martínez-Cañas 2014; Verquer, Beehr, and Wagner 2003). Supervisory abuse clashes with the ideal values of the public sector and is beyond the expectations of highly P-O-fitted subordinates. Cognitive dissonance strengthens the negative influences of abusive supervision as abusive supervisors create an unfair and hostile work environment and express damaging and adverse values to public servants with a high P-O fit. As highly P-O-fitted public servants uphold a shared value of ethical standards endorsed by public sector organizations, they are unlikely to tolerate supervisory abuse. They could feel more distress due to the difference in value from what they expect to have from their public sector organizations, in which ethical values are pertinent to ethical conduct (Gould-Williams, Mostafa, and Bottomley 2015; Thuy and Phinaitrup 2021). Additionally, abusive supervision generates mixed messages about organizational values that confuse public servants with high P-O fit. Therefore, they experience more damaging impacts from supervisory abuse. The message conveyed by supervisory abuse is more perceptible to public servants with high P-O fit because they desire to receive respect, trust, and support from their supervisors who represent the organization. Highly P-O-fitted public servants may find it more stressful to work with abusive supervisors. They will feel deprived from the lack of alignment between the values of the organization and the role modelling of the behaviours exemplified by supervisors, as expected in CDT principles. We hypothesize that high P-O fit could amplify the adverse relationship between abusive supervision and psychological distress. Meanwhile, P-O fit could heighten the positive relationship between psychological distress and turnover intention because abusive supervision is a stressful and unpleasant experience that does not reflect compatible values of desired and experienced treatment.

Hypothesis 4.

P-O fit intensifies the relationship between abusive supervision and psychological distress.

Hypothesis 5.

P-O fit intensifies the relationship between psychological distress and turnover intention.

Hypothesis 6.

P-O fit intensifies the relationship between abusive supervision and turnover intention.

Hypothesis 7.

P-O fit intensifies the relationship between abusive supervision and turnover intention via psychological distress.

Methodology

Sampling procedure and demographic information of respondents

Following the same recruitment strategy as Mayer et al. (2012), we targeted approximately 1,000 American respondents through Zoomerang.com, an online data collection service recently powered by SurveyMonkey. The panel of Americans registered with Zoomerang.com is highly comparable to those represented in the US Census in terms of age and annual household income, but females are slightly overrepresented in the panel (Mayer et al. 2012, 28). We excluded incomplete responses as well as responses from those who did not want to consent or did not self-identify as public servants. Based on post-hoc power analysis using G*power with small effect size (f ²=.15) and power = .80, the minimum sample size was 92 respondents. The final data set involving 468 complete and usable responses was accomplished in 2016 (a response rate of 43.9%), indicating sufficient power and effect size to produce accuracy and flexibility of estimates with five tested predictors including two interaction terms.

Of the respondents, more than three-quarters were at least 31 years old and above. The majority (94.2%) were full-time employed public servants. In this sample, female respondents were nearly two times more than male participants. Half of the respondents (50.2%) had more than 10 years of organizational tenure, and half (50.7%) were at junior and middle managerial levels. Nearly one-third reported working in a clerical or support role, while 43.4% worked in a professional or technical role in their organizations. Our sample contained 28.2% of respondents working at federal government agencies, 34.2% from state governments, and 37.6% from local governments. The demographic of the respondents was quite similar to that of the respondents who participated in the 2016 Federal Employee Viewpoint Survey. We performed statistical analysis and measurement model estimation using SPSS and AMOS ver28.

Measures

Abusive supervision

We measured this construct using a five-item scale from Mitchell and Ambrose (2007). The respondents were asked to indicate whether their supervisor had shown any abusive behaviours (e.g. my supervisor tells me I’m incompetent). The respondents rated the behaviours of their supervisors on a 7-point Likert scale ranging from ‘strongly disagree’ to ‘strongly agree’ (Cronbach’s alpha [α]=.96, Average Variance Extracted [AVE]=.82).

Psychological distress

We adopted a 10-item scale from Kessler et al. (2002) to measure psychological distress. We asked the participants to report how they often felt in the last 30 days (e.g. did you feel restless or fidgety?) on a 5-point Likert scale anchored from 1 (none of the time) to 5 (all of the time) (α=.94, AVE = .62).

Turnover intention

We asked the respondents to indicate whether they agreed with the turnover intention on a 3-item scale from Meyer, Allen, and Smith (1993) on a 7-point Likert scale. Sample items included ‘I frequently think about leaving my current employer’ (α = 91, AVE = .76).

P-O fit

We adopted a 4-item scale from Wright and Pandey (2008) to measure P-O fit using a 7-point Likert scale of agreement degree. Sample items included ‘my personal values match the organization’s values’ and ‘the organization’s values are a good fit with the things that I value’ (α = 97, AVE = .88).

Control variables

Using dummy codes, we controlled for age (2 = 18–25, 3 = 26–30, 4 = 31–40, 5 = 41–50, 6 = 51–60, 7= above 60), gender (1=male, 2=female), and organizational tenure (1= less than 12 months, 2 = 1–2 years, 3 = 3–5 years, 4 = 6–10 years, 5=more than 10 years). We also controlled for education (0=other, 1=some college, 2=diploma, 3=bachelors, 4=graduate degrees), job nature (1=clerical/support, 2=professional/technical, 3=service/maintenance, 4=manager/executive), managerial level (0=other, 1=senior management, 2=middle management, 3=junior management), and organizational type (0=other, 1=federal government agency, 2=military, 3=state government, 4=local government).

Measurement model estimation

The convergent validity of latent constructs was established as the values of AVE were above the threshold of .50 (Hair et al. 2010). We assess the discriminant validity of the five measurements based on the following checks. First, the results of the comparison of the goodness of fit between the hypothesized model (model 1) and the alternative models are presented in Table 1. Accordingly, model 1 had a better fit to the data (λ²[173] = 365.81, CFI = .98, TLI = .98, RMSEA = .05, SRMR = .04) compared to other alternative models. Second, the square root of the AVE value for each variable was higher than its correlation with any other variable (Fornell and Larcker 1981). Third, the value of the maximum shared variance (MSV) of each variable was smaller than its relative AVE (Hair et al. 2010). Finally, multicollinearity was not a significant issue in our study as the tolerance values (ranged from .768 to .941) were above the threshold of .10, and the VIF values ranged between 1.062 and 1.302 were below the threshold value of 3.3 (Hair et al. 2010). Given all the results, we contend that the latent constructs in this research had convergent and discriminant validity.

Table 1. Fit index comparison between hypothesized and alternative models.

Model		χ^2	df	CFI	TLI	RMSEA	SRMR	△χ^2
Model 1	Four-factor model (preferred model)	365.81	173	.98	.98	.05	.04
Model 2	Three-factor model (Abusive supervision + psychological distress, P-O fit, turnover intention)	1743.29	176	.84	.81	.14	.18	△χ^2(3) = 1377.48, p < 0.001
Model 3	Two-factor model (Abusive supervision + psychological distress + turnover intention, P-O fit)	2525.80	178	.76	.72	.17	.19	△χ^2(5) = 2159.99, p < 0.001
Model 4	Single-factor model	4914.59	179	.52	.43	.24	.24	△χ^2(6) = 4548.78, p < 0.001

Common method variance (CMV) checks

We conducted several tests to check for CMV (Podsakoff, MacKenzie, and Podsakoff 2012). First, Harman’s single factor test was performed by entering all the items of the latent variables using the principal components method with an analysis of the unrotated factor solution on SPSS ver28. This test revealed one single factor with eigenvalues of greater than one explained 39.74% of the variance in the exogenous and endogenous constructs. Second, we followed Lindell and Whitney (2001)to use the social desirability variable as the marker to check the potential of CMV by checking the difference between correlations of all constructs in the model, with and without the marker variable. The test showed the difference was 0.08 below the cut-off value of .20 (Lindell and Whitney 2001), indicating that the marker variable did not influence the correlations between our latent variables. Finally, the significant moderation effect in this study supported the confidence that CMV was not a major concern. Bozionelos and Simmering (2022, 198) noted that in complex models, such as the inclusion of moderation effect as theorized in this study, CMV is ‘unlikely to produce statistically significant relationships when they do not truly exist’.

Results

Table 2 reports the descriptive statistics and correlations between the variables of the study. As shown in Table 2, age was found to be negatively associated with psychological distress (b=−.22, p < .001) and turnover intention (b=−.10, p < .05). These findings indicate that older participants were less likely to have high psychological distress and turnover intention. Likewise, organizational tenure appeared to have a negative association with psychological distress (b=−.11, p < .05), indicating that respondents with a long experience with the organization were unlikely to report high psychological distress. Gender was found not to have any association with the latent variables. Education was found to have a negative association with abusive supervision and a positive relationship with person-organization fit (b=−.09 and .10 respectively, p < .05), whereas managerial level was positively related to person-organization fit (b = .11, p < .05).

Table 2. Descriptive statistics and inter-correlations between variables.

Variables	M	SD	CR	AVE	MSV	1	2	3	4	5	6	7	8	9	10	11
1. Age	-	-	-	-	-	1
2. Gender	-	-	-	-	-	−.06	1
3. Org tenure	-	-	-	-	-	.52***	−.07	1
4. Education	-	-	-	-	-	−.08	−.03	−.02	1
5. Job nature	-	-	-	-	-	.02	−.29***	.03	.22***	1
6. Managerial level	-	-	-	-	-	−.11*	−.07	.03	.19***	.15**	1
7. Org type	-	-	-	-	-	−.10*	.21***	−.12**	−.05	-.08	.06	1
8. Abusive supervision	1.80	1.31	.96	.82	.24	−.03	−.07	.03	−.09*	-.03	-.01	-.06	.91
9. Psychological distress	1.66	0.79	.94	.62	.17	−.22***	−.04	−.11*	−.04	-.01	.04	-.01	.44***	.79
10. Person-organization fit	4.00	1.53	.97	.88	.13	.02	.01	−.06	.10*	.03	.11*	.04	-.18***	-.10*	.94
11. Turnover intention	2.72	1.83	.91	.76	.24	−.10*	.05	.02	.03	-.02	-.02	.03	.51***	.38***	-.39***	.87

Note: N = 468, *p<.05, ***p<.01, ***p<.001. Bold and italic numbers are the square root values of AVE. Mean and Standard deviance of control variables are not reported due to the use of dummy codes.

Prior testing the hypotheses, we performed multivariate linear regression to examine the direct relationship between abusive supervision and turnover intention without and with control variables (please see Table 3). We then utilized Models 4 and 59 in the Process Macro on SPSS ver 28 (Hayes 2018) to test the hypotheses regarding the relationships among abusive supervision, psychological distress, and turnover intention (Model 4) prior to the introduction of the moderating effects of P-O fit (model 59). Table 4 reports the findings for hypothesis testing. Specifically, Model 4 in Table 4 shows that abusive supervision is positively and directly related to psychological distress (β=.26, p < .001) and turnover intention (β=.62, p < .001). Psychological distress was found to be positively and directly associated with turnover intention (β=.44, p < .001). A 95% confidence interval (CI) based on 10,000 bootstrap samples did not include zero, showing a relationship between abusive supervision and turnover intention through psychological distress (effect=.11, bootstrapped standard errors = .03, 95%CI[.06, .18]). Therefore, hypotheses 1, 2 and 3 were supported.

Table 3. Results of the direct relationship between abusive supervision and turnover intention.

Variables	Turnover intention
	β	SE	t	β	SE	t
Constant	1.42	.12	11.492***	.82	.54	1.52
Age	-	-	-	−.16	.07	−2.26*
Gender	-	-	-	.29	.16	1.78
Org tenure	-	-	-	.10	.07	1.46
Education	-	-	-	.12	.06	1.90
Job nature	-	-	-	.02	.08	.30
Managerial level	-	-	-	−.07	.06	−1.13
Org type	-	-	-	.08	.06	1.30
AS	.72	.06	12.92***	.73	.06	13.146***
	R^2=.26, F(1, 466) = 166.83***			R^2=.28, F(1, 459) = 172.83***

Note: N = 468, *p<.05, **p<.01, ***p<.001. AS=abusive supervision, PDS=psychological distress, POF=person-organization fit. Unstandardized coefficients are reported.

Table 4. Results for hypothesis testing.

	Model 4 on Process Macro for relationship among abusive supervision, psychological distress, and turnover intention								Model 59 on Process Macro for conditional effects of abusive supervision on turnover intention via psychological distress
	Psychological distress				Turnover intention				Psychological distress				Turnover intention
Variables	β	SE	t	95%CI	β	SE	t	95%CI	β	SE	t	95%CI	β	SE	t	95%CI
Age	-.12	.03	-3.91***	-.18, -.06	-.10	.07	-1.49	-.24, .03	-.11	.03	-3.75***	-.17, -.05	-.04	.06	-.63	-.17, .09
Gender	-.03	.07	-.40	-.17, .11	.30	.16	1.90	-.01, .61	-.02	.07	-.27	-.16, .12	.29	.15	2.01*	.01, .58
Org tenure	-.02	.03	-.50	-.08, .04	.11	.07	1.56	-.03, .24	-.02	.03	-.52	-.08, .04	.05	.06	.05	-.07, .18
Education	-.01	.03	-.45	-.07, .04	.12	.06	2.03*	.00, .24	-.01	.03	-.43	-.06, .04	.15	.06	2.72**	.04, .26
Job nature	-.00	.03	-.09	-.07, .06	.02	.08	.32	-.12, .17	.00	.03	.09	-.06, .07	.04	.07	.50	-.10, .17
Managerial level	.02	.03	.68	-.03, .07	-.07	.06	-1.28	-.18, .04	.01	.03	.50	-.04, .06	-.03	.05	-.63	-.14, .07
Org type	-.00	.03	-.17	-.06, .05	.08	.06	1.36	-.04, .20	-.00	.03	-.08	-.06, .05	.10	.06	1.79	-.01, .21
AS	.26	.02	10.31***	.21, .30	.62	.06	10.20***	.50, .74	.27	.03	10.57***	.22, .32	.56	.06	9.36***	.44, .68
PDS					.44	.10	4.33***	.24, .65					.42	.10	4.30***	.23, .61
POF									-.01	.02	-.46	-.05, .03	-.36	.04	-8.11***	-.44, -.27
Effect of AS on Turnover intention via PDS
Effect=.11, bootstrapped SE=.03, bootstrapped 95%CI=.06, .18
AS*POF									.04	.01	3.04**	.02, .07	.02	.03	.69	-.04, .09
PDS*POF													.13	.06	2.27*	.02, .25
Constant	1.91	.24	7.99***	1.44, 2.38	-.03	.56	-.06	-1.14, 1.07	.67	.22	2.94**	.22, 1.12	1.57	.48	3.29**	.63, 2.50
	R^2=.23, F(8, 459)=17.29***				R^2=.32, F(9, 458)=23.53***				R^2=.25, F(10, 457)=15.02***				R^2=.42, F(12, 455)=27.07***
Conditional effects of AS on PDS at values of POF
-1SD									.20	.03	6.89***	.15, .26
Mean									.27	.03	10.57***	.22, .32
+1SD									.34	.04	9.08***	.26, .41
Conditional effects of PDS on Turnover intention at values of POF
-1SD													.21	.13	1.70	-.03, .46
Mean													.42	.10	4.30***	.23, .61
+1SD													.62	.14	4.48***	.35, .89
Conditional effects of AS on Turnover intention via PDS
-1SD													.04	.03	-	-.01, .12
Mean													.11	.03	-	.05, .18
+1SD													.21	.05	-	.11, .31

Note: N=468, *p<.05, **p<.01, ***p<.001. AS=abusive supervision, PDS=psychological distress, POF=person-organization fit. Unstandardized coefficients are reported.

Based on Model 59, we found a positive association between abusive supervision and psychological distress (β=.27, p < .001). The moderation effect of P-O fit on the relationship between abusive supervision and psychological distress was significant (β=.04, p < .01). We then plotted the interaction effect in Figure 2 using one standard deviation above and below the mean of P-O fit (Aiken and West 1991). As can be seen, regardless of P-O fit, public servants in a more abusive relationship with direct supervisors tended to experience high psychological distress. Figure 2 indicated that the positive association between abusive supervision and psychological distress was strengthened when P-O fit was high (b = .34, se = .04, t = 9.08, p < .001, 95%CI[.26, .41]). This finding supported Hypothesis 4.

Figure 2. Simple slopes of abusive supervision on psychological distress at different levels of P-O fit.

According to Hypothesis 5, we expected a moderation effect of P-O fit on the relationship between psychological distress and turnover intention. We found that psychological distress was only significantly related to turnover intention when P-O fit increased, indicating that high P-O fit strengthened the positive relationship between psychological distress and turnover intention (b = .62, se = .14, t = 4.48, p < .001, 95%CI[.35, .89]), partially supporting Hypothesis 5. Figure 3 presents the simple plot for the interaction between psychological distress and P-O fit influencing turnover intention, indicating that the slope representing the association between psychological distress and turnover intention was sharper at a high P-O fit level.

Figure 3. Simple slopes of psychological distress on turnover intention at different levels of P-O fit.

The interaction between abusive supervision and P-O fit on turnover intention was not statistically significant (b = .02, p = .49), rejecting Hypothesis 6. Finally, a 95% confidence interval (CI) based on 10,000 bootstrap samples did not include zero, showing the relationship between abusive supervision and turnover intention through psychological distress is only statistically significant when P-O fit was high (effect=.21, bootstrapped standard errors = .05, 95%CI[.11, .31]). Therefore, Hypothesis 7 was partially supported. Overall, the R² of models in our study are greater than the thresholds for small (R²=.02) and medium effects (R²=.13), indicating that our models had a medium effect suggested by (Cohen 1988, ll. 413–414).

Discussion

This research integrates a comprehensive model of SCT to explore the mechanisms underlying the impact of abusive supervision on public servants. Overall, abusive supervision was positively related to turnover intention (Hypothesis 1). The findings also support the relationships: abusive supervision-psychological distress (Hypothesis 2) and psychological distress-turnover intention (Hypothesis 3). We also found P-O fit intensified the impact of abusive supervision on psychological distress (Hypothesis 4), but only magnified the relationship between psychological distress and turnover intention (Hypothesis 5) when P-O fit is high. We did not find significant evidence to support Hypothesis 6 that proposed P-O fit intensified the relationship between abusive supervision and turnover intention. Finally, we found a partial support for Hypothesis 7 as only high P-O fit moderated the relationship between abusive supervision and turnover intention through psychological distress. These findings offer important theoretical and managerial implications for public administration.

Theoretical implications

Our first theoretical implication pertains to the limited body of scholarly research that investigates how the dark side of leadership influences turnover intention among public servants (see Backhaus and Vogel 2022; Hassan 2019; Hur and Abner 2023). Our theorized model was derived from SCT perspective, which portrays the influences of environmental factors on how individuals perceive, interpret, and navigate the workplace (Bandura 1986, 2001). We found that the harm to public servants with abusive supervision involves the experience of psychological distress and turnover intention. Public servants cognitively see abusive supervisors as detrimentally threatening the individual safety and protection that they expect from the organization. Also, abusive supervision is an interpersonal stressor to self-worth and personal growth (Jiang et al. 2021; Tepper, Simon, and Park 2017) as public servants seek positive interactions with supervisors to stay motivated, engaged, committed to the organization, and contribute to the delivery of public value (Bottomley et al. 2016; Lagowska, Sobral, and Tavares 2022; Park and Rainey 2008; Tavares, Sobral, and Wright 2021). Although public sector organizations endorse ethical conduct, supervisory abuse delivers contradictory messages to public servants. It makes them feel psychologically distressed while interacting with abusive supervisors. Therefore, leaving the organization might be a feasible coping behaviour for minimizing negative cognitions and psychological discomfort caused by abusive supervision while having uncomfortable and unfavourable perceptions of the organization (Rice, Taylor, and Forrester 2020; Saleem et al. 2021). Their significant associations in this study align with the tenets of SCT that describe the relationships between work environment and personal cognitive factors in explaining human functioning. As the nature of public administration could be a fertile ground for abusive supervision, our study shows that supervisory abuse can produce detrimental consequences, including mental health problems (e.g. psychological distress) in employee retention, which is lacking in the extant literature. It will be more substantial to consider other damaging outcomes of the dark side of leadership in future research to develop a more holistic understanding of prevention strategies to protect the health and safety of public servants.

Second, we add further nuance to our understanding of the link abusive supervision – psychological distress – turnover intention by examining the moderating role of P-O fit. Our research demonstrates that highly P-O-fitted public servants tend to feel highly distressed as supervisory abuse increases a meaningful discrepancy. Also, we found that a high P-O fit heightens the motivation to leave the organization when the work situation becomes stressful and harmful due to the aversive experience with supervisory abuse. Our findings compete with the notion that a high P-O fit discourages turnover (Gould-Williams, Mostafa, and Bottomley 2015; Jin, McDonald, and Park 2018; Moynihan and Pandey 2008; Tavares, Sobral, and Wright 2021). Indeed, the integrated SCT and CDT aspects provide a substantial foundation to exemplify the potential that public servants with a high P-O fit are prone to psychological distress and turnover intention as they are more aware of and susceptible to the violation of value congruence caused by abusive supervisors. Based on our research, it seems that P-O fit is a central part of human functioning that directs their views of the organization through the expression of supervisors who are symbolic characters of the organization. Our study offers new insights into the traditional approach to examining determinants of turnover intention in PA research. These findings suggest new avenues for future PA research by integrating theoretical frameworks to advance the body knowledge of determinants of turnover intention among public servants while considering individual variations. Doing so provides new frameworks to investigate the interplay of workplace situations, including leadership behaviours, personal factors, and psychological cognitive processes.

Lastly, our study advances SCT by integrating CDT to explain the moderating role of P-O fit in the relationship among abusive supervision-psychological distress-turnover intention. Although SCT has argued for the influence of environmental and personal factors in explaining human behaviours, it is unclear how individuals react differently to such triggers. As SCT is limited in explaining whether public servants react to abusive supervision similarly, we integrated CDT into SCT to explain the circumstances in which public servants with unique P-O fit degrees may differentially respond to abusive supervision. CDT portrays that when individuals perceive dissimilarities or discrepancies between their values and those of the organization expressed by supervisory forms, they could engage in coping behaviours to resolve cognitive dissonance (Festinger 1957). As we expected in Hypothesis 4, we found evidence of the moderating role of P-O fit in the negative relationship between abusive supervision and psychological distress. Public servants with high P-O fit tend to possess more negative cognitions and discomfort while interacting with abusive supervisors. Also, we found that the slope representing the positive association between psychological distress and turnover intention is significantly increased and sharper when the P-O fit is high, which supports our Hypothesis 5. These findings reinforce the theoretical CDT assertions and empirical findings about the role of P-O fit in varying employee experiences and reactions to specific work situations (Kristof-Brown, Treviño, and Harrison 2005; Mostafa et al. 2022; Ruiz-Palomino and Martínez-Cañas 2014; Verquer, Beehr, and Wagner 2003). The present study furthers the SCT literature in PA by showing how an environmental factor (abusive supervision) influences a behavioural factor (turnover intention) via an intrapersonal factor (psychological distress); as well as how public servants with different levels of P-O fit respond to such environmental stimuli. This research demonstrates new research efforts to integrate SCT with other theories (i.e. CDT) to offer better understanding of public servants’ behaviours.

Managerial implications

Our findings provide important implications for public sector organizations. We found that abusive supervision is detrimental to public servants’ mental health, which fosters their turnover intention. Our study pertains to the critical retention matter of the public sector that highlights the prominent value of public servants with a high P-O fit. High P-O fit public servants share the motivation to serve the greater good with the public sector organizations and often show better work performance, contributing to organizational success in achieving the mission of serving the public (Gould-Williams, Mostafa, and Bottomley 2015; Jin, McDonald, and Park 2018; Thuy and Phinaitrup 2021). The noteworthy implication of our study is that abusive supervisors greatly influence public servants with a high P-O fit. As such, abusive supervision can be especially harmful to public sector organizational effectiveness since it has more damaging impacts on highly P-O-fitted public servants. It is too confident to believe that public servants with a high P-O fit always stay and contribute to the organization as they can overcome stressful demands. Our study shows that letting highly P-O-fitted public servants experience supervisory abuse and a toxic work environment undermines their affinity with the organization and increases their likelihood of turnover. These findings provide crucial information for public sector management: an organization’s supervisors are core in managing distress and retaining public servants, especially those with a high P-O fit. Public sector organizations need to consider the potential breach in the organizational values caused by the expression of abusive supervision. To ensure public servants are best placed to deliver public values and are committed to the organization, minimizing the abusive behaviours of supervisors is critical. Initiatives that public sector organizations should consider include training supervisors in ethical leadership and establishing an ethical culture that reinforces positive leadership behaviours and reduces abusive behaviours among supervisors (Aryee et al. 2007). Public sector organizations need to vigilantly address supervisory abuse amongst public servants.

Limitations and future research directions

We acknowledge some limitations in this study. First, the potential of CMV could arise from the cross-sectional design (Podsakoff, Scott, and Podsakoff 2012). Nonetheless, such an approach may be adequate for studies aiming to provide initial evidence for an understudied phenomenon (Spector 2019). In this instance, as research has documented evidence on the consequences of abusive supervision (Fischer et al. 2021), we still know little about the boundary conditions for its flow-on impact (Smith et al. 2022). Based on the SCT, we primarily explored whom abusive supervision influences most, contributing to the literature by testing a condition that arises when public servants with different degrees of P-O fit experience abusive supervision and by revealing the conditional motives of turnover intention. While we are aware of the inability to conclude the causality of the variables, a cross-sectional design remains relevant to ascertain initial evidence of the conditional effects of abusive supervision on psychological distress and turnover intention that is not currently comprehended. Also, the CMV checks and the incorporation of an interaction term in the moderated mediation model gave us assurance that CMV would not significantly bias our findings (Bozionelos and Simmering 2022). Despite several CMV tests performed in this study, future research needs to utilize other remedies and additional statistical checks, such as non-response bias techniques, to alleviate the potential of CMV. We also suggest a longitudinal design, research with multiple sources of respondents, or an experimental approach to reduce the potential of enduring CMV arising from characteristics of respondents in studies using time separation or several sources of respondents as well as verify the causality of variables in this study (Podsakoff, Scott, and Podsakoff 2012; Spector 2019).

Second, we observed another limitation regarding the small interaction power of P-O fit to detect whether the associations of intended variables changed when conditioned on P-O fit (i.e. 16% detectable effect size for the conditioned relationship between abusive supervision and psychological distress and 17% detecting the conditioned association between psychological distress and turnover intention). Although interaction effects are usually quite small (Baranger et al., n.d..), future studies should consider a larger sample size that affects power analysis and plan for a lessened interaction impacts that are at least a third or half of the size of the main effects of interest (Baranger et al., n.d.; Gelman and Carlin 2014).

Third, we only examined the intention to quit and provided initial evidence of the possibility of leaving their organizations when highly P-O fitted public servants perceive abusive supervision. Some studies maintain that self-reported turnover intention does not always warrant the actual turnover behaviour because actual turnover behaviour is a more difficult decision and depends on several aspects of financial, personal, and life situations (Cohen, Blake, and Goodman 2016). Nonetheless, existing literature generally agree that the turnover intention could reasonably represent actual turnover (Cho and Lewis 2012; Stater and Stater 2019; Van der Heijden et al. 2018). Future research could include both turnover intention and actual turnover (Cohen, Blake, and Goodman 2016; Van der Heijden et al. 2018) to strengthen the mechanisms we explored in this study.

Finally, this study was undertaken with a sample of US public servants prior to the COVID-19 pandemic. As such, we recommend an extension to other public sector contexts to generalize the research findings and discover additional potential boundary conditions (e.g. cultural values) for the consequences of abusive supervision. Also, the literature (e.g. Dzigbede, Gehl, and Willoughby 2020; Schuster et al. 2020) has highlighted the significant changes in work design and interpersonal relationships during the pandemic; therefore, a larger sample size of public servants in the post-pandemic period is needed to validate and extend our research findings. Such research would be insightful in guarding the epistemic diversity of the scholarship on abusive supervision.

Conclusion

Although highly P-O fitted public servants are critical for the effectiveness of public sector organizations, very little is known about why such beneficial individuals decide to quit. Our findings demonstrate that abusive supervision impacts public servants’ psychological distress and turnover intention, and this detrimental effect is even more severe for those with a high P-O fit. Our research is particularly important since it provides insights into destructive leadership that has not been widely researched in the public sector. Furthermore, we show the cost associated with abusive supervision is extremely high since it can turn away important human assets of public sector organizations – those with a high P-O fit. It is vital that public sector organizations beware and have appropriate measures to deal with abusive supervision.

Disclosure statement

No potential conflict of interest was reported by the author(s).