How Ethical Leadership Impacts Deviant Normative Conduct? The Role of Trait Affect, Voice Behaviour, and Social Support

ABSTRACT

Objective: This study developed and validated ethical leadership and deviant normative conduct model through mediating variables of trait affect as a psychological process and subordinate voice behavior. In addition, we also tested the moderating influence of social support in the association between subordinate voice behavior and deviant normative conduct.

Method: The sample of the study was comprised of 258 pairs of employees and their direct leaders from six different organizations in China. We performed a structural equation model and hierarchical linear tests to check the hypothesized relationships.

Results: The results indicated that ethical leadership has a positive impact on positive affect and negative impact on negative affect. Positive affect and negative affect have a differential positive and negative influence on subordinate voice behavior that, as a result, has a positive impact on organizational citizenship behavior and negative influence on counterproductive work behavior. This study also confirmed the role of social support as an enhancer and suppressor of deviant normative conduct when social support interacts with subordinate voice behavior.

Conclusion: Our study uniquely applied social learning and conservation of resource theories to ethical leadership and normatively (un)desirable conduct at work, which have important implications for theory and practice.

KEY POINTS

What is already known about this topic:

(1) Ethical leadership has a positively direct effect on employee normatively desirable conduct.

(2) Ethical leadership has a negatively direct effect on employee normatively undesirable conduct.

(3) A single mediating mechanism exists until now in the relationship between ethical leadership and normative conduct, which are explored individually.

What this topic adds:

(1) Distinctively integrates and empirically validates ethical leadership and deviant normatively (un)desirable conduct of subordinates at work relationships, simultaneously.

(2) Synthesized psychological processes (operationalized as trait affect) and challenging behaviour (operationalized as voice behaviour) as underlying serial mechanisms.

(3) Recognized the role of social support as a crucial enhancer and suppressor simultaneously.

KEYWORDS:

• Ethical leadership
• trait affect
• voice behaviour
• social support
• deviant normative conduct

Introduction

How ethical leadership (EL) impacts normative conduct? The literature on EL consistently indicated that it has a positive influence on normatively appropriate conduct at work, such as organizational citizenship behaviour (OCB) (Gerpott et al., 2019; Kacmar et al., 2011; Lu, 2014; Mayer et al., 2009). However, studies also have shown that EL has a negative effect on employees’ undesirable conduct, such as counterproductive work behaviour (CWB) or employee deviance (Bush et al., 2020; Mayer et al., 2009; Schaubroeck et al., 2012). OCB is considered normatively appropriate conduct as it contributes positively to organizational effectiveness, while CWB is deemed to be normatively inappropriate conduct as it violates social and/or organizational norms (Fox et al., 2001). Brown et al. (2005) referred EL to a leadership style that shows normatively desirable behaviour through interpersonal relationships and individual acts, and the promotion of such behaviour to through two-way interaction of subordinates, decision-making, and reinforcement.

Nonetheless, there is enough evidence on the impact of EL on normatively (un)desirable conduct at work (Bedi et al., 2016; Khokhar & Zia-ur-Rehman, 2017; Ng & Feldman, 2015; Shareef & Atan, 2019). Yet, the knowledge of the mechanism between the relationship of EL and normative (un)desirable conduct is considered at its growing stage. Even though prior research reported that EL has a significant association with OCB and CWB, yet, with few exceptions (Peng & Kim, 2020; C. Yang et al., 2016) research on EL and employee’s conduct at work have not reported more than one or serial mechanisms. Examining a single mediating mechanism usually leads to the issue of “specious mediation,” which is a mediation that seems to channel an influence of the mediator; however, in reality, it does not do so (Fischer et al., 2017; Tourigny et al., 2019). Although, a recent meta-analysis found converging evidence of EL positive and detrimental effects on OCB and CWB (Peng & Kim, 2020). However, this limited literature has not achieved unanimity on how EL impacts OCB and CWB simultaneously.

Based on a systemic review, this study considers two primary serial mechanisms to answer this “how” question. The first is the psychological state, which is regarded as a state of a person’s cognitive processes in the field of psychology (Chen, 2003). It is argued that when subordinates understand that the leaders behaving ethically, then their cognitive state is positive, results in employee’s productive behaviour (Chen & Hou, 2016; Hoyt et al., 2013). However, if employees perceive unethical behaviour, their cognitive state is inclined towards negativity (Ariely, 2012), which results in counterproductive behaviour (Detert & Burris, 2007; Hoyt et al., 2013). Prior studies considered various variables as an underlying mediating mechanism in the relationship of EL and OCB, such as ethical cognition (Resick et al., 2013), self-efficacy and organizational identification (Walumbwa et al., 2011), person-organization fit and psychological safety (Ruiz-Palomino & Martínez-Cañas, 2014; Walumbwa & Schaubroeck, 2009), interpersonal trust (Lu, 2014), and leader-member exchange, ethical culture, organizational identification and trust in a leader (Peng & Kim, 2020). However, regardless of prior studies’ suggestion, consideration of psychological processes as a causal mechanism in the relationship between EL and deviant normative conduct (Peng & Kim, 2020) has been neglected to examine. Hence, it raises the question of how well more than one psychological state as underlying mechanisms explains the relationships between EL and deviant normative conduct (OCB & CWB). We consider trait affect as a psychological state in the present study.

In response to a recent meta-analysis call (Peng & Kim, 2020) that psychological processes, such as positive affect (PA) and negative affect (NA), need be tested as mechanisms between EL and normative conduct. PA and NA are the two dimensions of trait affect (Y. Liu et al., 2020). Trait affect is referred to as employee emotional feeling and experience at work (Bang & Reio, 2017). PA indicates the extent to which employee feels excited, energetic, and vigilant, while NA is a distinct element of un-gratifying engagements and personal unhappiness, which comprises of various aversive mood conditions, including irritation, contempt, hatred, guilt, distress, and nervousness (Watson et al., 1988). Hence, we assume that EL is positively related to PA (H1), and EL is negatively related to NA (H2).

An ethical leader conveys high moral standards and inspires subordinates to voice their feelings and recommendations about issues related to ethical conduct at work and other work-related processes (Islam et al., 2019; Walumbwa & Schaubroeck, 2009; Zhu & Akhtar, 2019), such as OCB and corporate social responsibility (Tourigny et al., 2019). Social learning theory (SLT; Albert Bandura, 1977) suggests that people shape their conduct on those whom they trust and respect. Furthermore, the norms and behaviours of EL encourage employees to voice their opinions and ideas (Chen & Hou, 2016). Hence, it is argued that EL is an important predictor of employee voice. Although prior studies have offered valuable insights into the somewhat stable determinant of voice, the phenomenon that employees psychological state describes the association between EL and employees voice behaviour (VB) as an underlying mechanism has been mainly neglected. In response to call (Detert & Burris, 2007), that psychological mechanism should be studied in the association between leadership behaviour and subordinates’ voice; the study at hand considers psychological state (trait affect) is an underlying mechanism in the association between EL and VB. VB is referred to as behaviour portraying a constructive challenge aimed to improve a state of affairs (W. Liu et al., 2010; Van Dyne & LePine, 1998). Thus, we hypothesize that PA is positively related to VB (H3), and NA is negatively related to VB (H4).

The existing literature indicates that VB is valuable in the workplace (e.g., Chen & Hou, 2016; Maynes & Podsakoff, 2014). Voice is a challenge-oriented behaviour that involves expressing opinions affecting the work environment and contradictory with the status quo (Li & Sun, 2015; McClean et al., 2013). Conservation of resource (COR; Hobfoll, 1989, 2002) theory states that people have both a learned and an innate drive to protect, conserve, foster, and create the quantity and quality of their resources. Corresponding to COR, the resources that individuals strive to undertake, preserve, and protect are deemed useful objectives, abilities, conditions, or personal traits. Thus, we presume that COR theory may be especially valuable in justifying how the voice of subordinate is associated with employees’ deviant normative conduct. Moreover, as VB is considered as an important citizenship behaviour indicator (LePine et al., 2002), subordinates can dynamically raise voice to acquire a constructive response from their supervisors and colleagues (Fuller et al., 2007). However, when people realize the risk linked with VB, they may even raise their voice since of their devotion to supporting their moral values or beliefs (Gentile, 2012; Painter et al., 2019). Mackenzie et al. (2011) suggest that VB employees improve work effects when workers are driven by EL. Due to the ethical decision-making procedures and results, leaders determine particular ethical principles for their followers by driving appropriate conduct (in line with the principles) and discouraging misbehaviour (inconsistent with and violate the principles) (Treviño et al., 2006). Thus, we assume that VB at the same time can lead to employees’ normatively appropriate conduct (i.e., OCB), and inappropriate conduct (CWB). Hence, this study includes VB is the second level mediation after psychological states in the relationship EL and deviant normative conduct. Accordingly, we propose that VB is positively related to OCB (H5), and VB is negatively related to CWB (H6).

This study also aims to check the moderating impact of social support in the relationships between subordinate VB and deviant normative conduct (OCB and CWB). Social support indicates the degree to which a task presents an opportunity for guidance and support from others (Morgeson & Humphrey, 2006). We consider that social support will enhance OCB when social support interacts with subordinates VB. However, social support will suppress the detrimental effect of CWB when social support interacts with subordinates VB. Accordingly, we assume that social support strengthens the positive association between VB and OCB (H7), and social support weakens the negative association between VB and CWB (H8). Based on SLT of Bandura’s (Bandura, 1969, 1986) and COR theory (Hobfoll, 1989, 2002), yet, less emphasis has been given to analyse the psychological processes (Detert & Burris, 2007; Peng & Kim, 2020), and employees VB (Van Dyne & LePine, 1998). To tackle this limitation in the literature, the paper at hand offers a model that all together reflects the social learning and COR theories linkages through which EL has been hypothesized to have a deviant impact on normatively (un)desirable conduct at work (see Figure 1).

Figure 1. Ethical leadership and deviant normative conduct at work model showing hypothesize relationships

Methods

Procedure

The data of this study were gathered from six distinct companies across China through a survey instrument. The respondents were employees (subordinates) and their direct managers (leaders). We compile a list of potential respondents from the selected companies. These respondents were identified via local human resource (HR) forums, including company HR executives. We described the study aim and data gathering process, and all the HR executives approved to take part in the survey. Accordingly, they provided the email and WeChat IDs of the leaders/managers and their direct subordinates. To circulate the questionnaire and gather data, 312 employees were randomly selected from these six companies. Next, we split the respondents into two groups; the first group was comprised of employees, and the second group was including the selected employees’ immediate supervisors. Accordingly, we divided our questionnaire into two surveys; the first survey was about EL, PA, NA, and social support (these variables comprised of employees rated items), and the second survey was about employees’ VB, OCB, and CWB (consist of supervisors rated items). Furthermore, we collected the data through an online database; hence, we created two online links that included relevant items about corresponding variables.

We sent the first survey instrument link to the email’s addresses and WeChat ID of the employees, and the second survey instrument link was dispatched to relevant employees’ immediate supervisors. We received 268 useful responses from the first survey instrument link; however, 285 were received from the second survey instrument link. We matched the relevant employee-supervisor data through the email addresses and WeChat ID of the corresponding organizations that we already saved in a file. We successfully matched 258 responses from the employees and their immediate supervisor’s data files that were useful to be used for the analysis purpose.

Participants

The final dyadic sample (N = 258; supervisor-employee) was comprised of 152 male supervisors (58.91%) and 169 female subordinate employees (65.50%). The average age of supervisors was 35.29 years (SD = 14.67), while the employee’s average age was 27.82 years (SD = 9.13). The results of the demographic variables reveal that 121 (46.9%) supervisors have a master-level university degree, while 139 (53.88%) employees were college graduates. The supervisor had on average 5.6 years of tenure, while their subordinates had on average 3.9 years of tenure in the current position.

Measures

On the whole, a 7-point Likert scale was used to measure all the constructs of our study that was ranging to 7 = completely agree from 1 = completely disagree. However, in the case of CWB, respondents were asked to rate the items on a 7-point Likert scale that was ranging to 7 = often from 1 = never.

Ethical leadership

To assess EL (α = .86), we utilized the 10-item EL scale developed by Brown et al. (2005), which is the frequently used instrument in the research of management studies and operationalizes this study’s conceptual description of EL. An example item includes “My leader disciplines employees who violate ethical standards.”

Trait affect

Trait affect was measure through 20 items developed by Watson et al. (1988). Trait affect has two dimensions, i.e., PA (α = .82) and NA (α = .77). Each dimension was measure through 10 items (e.g., “determined” was representing PA and NA was by “distressed”).

Voice behaviour

Employee VB (α = .89) was measured through six items voice scale developed by Van Dyne and LePine (1998). Li and Sun (2015) also used employee VB scale in the Chinese context. An example item includes “This employee develops and makes recommendations concerning issues that affect our work group.”

Social support

Six items construct of social support (α = .81) from the scale of Work Design Questionnaire (WDQ) was adopted from Morgeson and Humphrey (2006), who developed this construct to assess and analyse the nature of work and job design. A sample item is “My supervisor is concerned about the welfare of the people that work for him/her.”

Normative conduct

Normative conduct constructs were measuring the social norms that depict what the typical or usual thing to do as descriptive norms at work is. Overall, two types of deviant normative conduct were encompassing the construct, i.e., OCB (α = .92) and CWB (α = .88). OCBI (α = .89) (OCB individual) and OCBO (α = .95) (OCB organizational) were representing the overall OCB construct. The items for these constructs were adopted from the study of Williams and Anderson (1991). Each dimension was measured through seven items. Sample items include “This employee assists me with my work (when not asked)” (OCBI), and “This employee conserves and protects organizational property” (OCBO). However, CWB was measured through 19-items adopted from the study of Bennett and Robinson (2000), who developed the overall CWB construct represented by two dimensions, i.e., interpersonal deviance (α = .87) and organizational deviance (α = .91). Sample items include “Cursed at someone at work” (interpersonal deviance), and “Come in late to work without permission” (organizational deviance).

Control variables

We consider gender, age, education, and tenure as control variables. However, gender and age, in the case of both supervisor and subordinate, were not significantly correlated with any main variables of the study. Hence, they were removed from the analysis. Supervisor education and subordinate tenure were significantly correlated with the OCB and CWB. Therefore, we decided to include supervisor education and subordinate tenure as control variables. Moreover, supervisor tenure and subordinate education were not significantly correlated with normative conduct variables; hence, we excluded these demographic variables from the analysis.

Analyses

IBM SPSS Amos was used to perform a structural equation model (SEM) tests to check and approve the hypothesized relationships presented in Figure 1. Following Anderson and Gerbing’s (1988) suggestions, our study used two steps method to examine the construct and test hypothesized relationships. First, confirmatory factor analyses (CFA) was performed to confirm the validity of the latent variables and measurement model. Second, to estimate the measurement model, we performed discriminant and convergent validities tests. Following Hu and Bentler (1999) suggestions several fitness indicators were used to check the model fitness, for example, Chi-square (χ²) goodness of fit index (GFI), comparative-fit-index (CFI), root mean square error of approximation (RMSEA), and standardized root mean square (SRMR). These indicators indicate the overall better fit of the study hypothesized model in comparison to the null or independent model. The association between the observed variables is supposed to be null, i.e., zero in the model known as null or independent.

Since the data was collected from two different sources, i.e., the first source was the supervisors/managers and the second sources were the subordinates/employees, according to the recommendations of Podsakoff et al. (2003) common method bias should not be considered a serious issue related to our study. However, according to Podsakoff and Organ (1986) indications about the self-reported nature of the scale usage, common method variance (CMV) likely to play a role related to data biases, i.e., a single factor will emerge from the factor analyses, and the majority of the covariance may be resultant for one general factor. Hence, following Harman’s single factor test method, we performed exploratory factor analysis (EFA) to explore the issue. The EFA scores indicate that the first factor explains only 21% of the variance. Therefore, no single factor was reported that claim majority of the shared variance, which shows that CMV is not a serious issue. Furthermore, we also performed a more sophisticated test of common latent factor method recommended by Podsakoff et al. (2003) in addition to the Harman test. Accordingly, after adding a common variable, and it was linked in the structural model to all exogenous and endogenous variables (observed). The common factor with and without values indicated insignificant variances between the models. Thus, the common latent factor test verified the results of the Harman test, reporting that CMV is not the issue related to our study.

To explore the multicollinearity issue among the studied variables, we performed collinearity diagnostics. Following Myers’s (1990) suggestion about variance inflation factor (VIF) and Menard’s (2002) suggestion relating to tolerance statistic values were considered as a standard to check multicollinearity issues. Accordingly, all VIF score was less than the standard value of 10 (Myers, 1990), and the tolerance statistic was well above than the standard value of .2 (Menard, 2002). Hence, we did not count the multicollinearity issue into consideration that takes an alarming posture associated with our study key variables.

We performed a one-way analysis of variance (ANOVA) with EL as a dependent variable, and the results indicated that between groups variance was not higher compared to within-group variance (F = .89, p = .59). As a result, the conceptualization of our study model at the individual level was supported. All the above analyses permit us to proceed to test the structural relationships of the hypothesized model. In addition to the SEM analyses, we performed hierarchical linear modelling to examine the moderating role of social support. To further explore the moderating influence of social support in the relationship between VB and deviant normative conduct at work, i.e., OCB and CWB, a test of the simple slope was conducted in addition to hierarchical linear modelling.

Results

The intercorrelations among the main variables are shown at Table 1, and the results illustrating that the most important variables of the study were significantly intercorrelated with one other. The mean (M), and standard deviation (SD) of the examined variables were reported as; EL (M = 5.01, SD = .84), PA (M = 4.83, SD = .74), NA (M = 3.34, SD = .69), VB (M = 4.98, SD = .91), social support (M = 5.22, SD = .97), OCB (M = 4.55, SD = .81), and CWB (M = 3.49, SD = .67). Next, we performed SEM analyses to test model fitness and structural relationships.

Table 1. Intercorrelations of variables

Variables		1	2	3	4	5	6	7	8	9
1	Supervisor education	1
2	Subordinate tenure	.01	1
3	Ethical leadership	.07*	.06	1
4	Positive affect	.00	.03	.31**	1
5	Negative affect	−.02	.01	−.24**	−.05	1
6	Voice behaviour	.01	.04	.11*	.43**	−.51**	1
7	Social support	.07*	.00	.05	−.03	−.06	.22**	1
8	OCB	.15**	−.07*	.22**	.12*	−.12*	.57**	.19**	1
9	CWB	−.08*	.11*	−.18**	−.09*	.22**	−.31**	−.25**	−.55**	1

*p < .05; **p < .0

The latent variables’ validity was determined by performing CFA. Individual variable CFA of first-order constructs were: EL (χ²/df = 1.19, GFI = .96, CFI = .98, RMSEA = .04, SRMR = .02), PA (χ²/df = 2.24, GFI = .93, CFI = .95, RMSEA = .05, SRMR = .03), NA (χ²/df = 2.21, GFI = .94, CFI = .96, RMSEA = .05, SRMR = .03), VB (χ²/df = 1.13, GFI = .98, CFI = .99, RMSEA = .02, SRMR = .01), social support (χ²/df = 2.28, GFI = .92, CFI = .93, RMSEA = .05, SRMR = .04). OCB was regarded as second-order construct comprises of two first-order dimensions (i.e., OCBI and OCBO) (χ²/df = 2.29, GFI = .94, CFI = .95, RMSEA = .05, SRMR = .04). In addition, CWB was also treated as a second-order construct comprises of two first-order dimensions (i.e., interpersonal deviance and organizational deviance) (χ²/df = 2.26, GFI = .96, CFI = .97, RMSEA = .04, SRMR = .02). The CFA of individuals variables shows that the values for GFI and CFI were well above as compared to the acceptable value, i.e., .90, and for SRMR, the values were below the required value of .05.

The analysis of the second-order (two-factor) model indicated better fit one-factor model (∆χ² = 899.23, p < .001). This result exhibited that the concept of considering OCB and CWB as distinct constructs are supported. The composite reliability of the examined variables was in-between .79 to .96, significantly above the required value of .60. Similarly, individual factor average variance extracted scores were in-between .54 to .65, higher as compared to the threshold value of .50, as suggested by Fornell and Larcker’s (1981). Thus, confirming our study construct convergent and discriminant validities. The adaptability test result of the measurement model indicated a good fit to the data (χ²/df = 2.87, GFI = .91, CFI = .94, RMSEA = .07, SRMR = .05). The structural modelling technique was employed to assess theorized relations. The hypothesized model fit indices indicate significant and better fit (χ²/df = 2.11, GFI = .97, CFI = .98, RMSEA = .04, and SRMR = .02), indicates that our hypothesized model has a good fit to the data. Furthermore, the recommendations of Anderson and Gerbing (1988) were followed to test five alternative models that were anticipated to be less fit to the data set. As anticipated, the model fit indexes shown in Table 2 indicate that the hypothesized model has a good fit for the data than the other five alternative models.

Table 2. Results of models fit indexes

Type of models tested	χ^2/df	GFI	CFI	SRMR	RMSEA
Hypothesized model	2.11	.97	.98	.02	.04
Alternative model 1: EL to VB (direct link)	3.48	.88	.89	.08	.09
Alternative model 2: EL to OCB (direct link)	3.72	.86	.87	.09	.10
Alternative model 3: EL to CWB (direct link)	3.99	.81	.83	.11	.14
Alternative model 4: deleted direct link from VB to OCB	4.34	.78	.80	.14	.15
Alternative model 5: deleted direct link from VB to CWB	4.89	.76	.77	.15	.16

Figure 2 shows the results of SEM analysis. All the path coefficient shown are based on standardized estimates. Hypothesis 1 (H1) states that EL is positively related to PA. The result reveals that the EL has a significantly positive influence on PA (β = .29, p < .01). Hence, H1 is supported. Hypothesis 2 (H2) asserts that EL is negatively related to NA. The result shows that EL has a significantly negative influence on NA (β = −.24, p < .01). Thus, H2 is supported. Hypothesis 3 (H3) declares that PA is positively related to VB. The result shows that leader PA has a significantly positive influence on subordinate VB (β = .31, p < .01). Hence, H3 is supported. Hypothesis 4 (H4) states that NA is negatively related to VB. The result indicated that leader NA has a significantly negative influence on subordinate VB (β = −.42, p < .01). Therefore, H4 is supported. Hypothesis 5 (H5) states that VB is positively related to OCB. The result indicated that the VB of subordinate has a significantly positive influence on OCB (β = .45, p < .01). Therefore, H5 is supported. Hypothesis 6 (H6) states that VB is negatively related to CWB. The result indicated that the VB of subordinate has a significantly negative influence on CWB (β = −.27, p < .01). So, H6 is supported.

Figure 2. Results of SEM analysis. All the paths reported standardized values of beta

Hypothesis 7 (H7) states that social support significantly interacts with VB to have a stronger influence on OCB. Table 3 results indicate that social support significantly interacted with VB to have a positive impact on OCB (β = .27, p < .01). Figure 3 reveals that VB at a higher level is related to higher OCB, VB is most likely to be more significant in effecting OCB when subordinates recognize social support. A simple slope test was also performed, which validated the above findings and indicated that VB has a stronger and positive impact on OCB when a subordinate receives a positive and higher level of social support (β = .37, t = 2.63, p < .01), than when such social support is low and negative (β = −.01, t = .06, p = .953). Therefore, H7 is supported.

Table 3. Results of hierarchical linear modelling

Variables	OCB			CWB
	Step 1	Step 2	Step 3	Step 1	Step 2	Step 3
Control variables
Supervisor education	.11 (.14)	.12 (.15)	.12 (.15)	−.01 (.07)	−.01 (.07)	−.03 (.08)
Subordinate tenure	−.06 (.09)	−.07 (.10)	−.07 (.11)	.08 (.10)	.08 (.11)	.09 (.12)
Predictors
Voice behaviour (VB)		.39 (.19)**	.18 (.22)*		−.45 (.23)**	−.22 (24)*
Social support (SS)		.04 (.09)	.09 (.12)		.01 (.10)	−.12 (.14)
Interaction
VB × SS			.27 (.15)**			.59 (.17)**
∆R^2		.08	.19		.12	.24

Note. Parentheses indicate standard errors. *p < .05; **p < .01.

Figure 3. Moderation of social support (SS) in the link between voice behaviour (VB) and OCB

Hypothesis 8 (H8) highlights that social support significantly interacts with VB to weaken CWB. Results shown at Table 3 confirms H8 postulation (β = .59, p < 0.01). Furthermore, Figure 4 shows that the effect of VB on CWB is lower when the social support intensity is at a higher level than at a lower level. Besides, the examination of simple slope analyses also confirmed that the VB has a negative and weaker effect on CWB when subordinate notices positive and a higher level of social support (β = .16, t = 1.21, p = .288), than when such social support is low and negative (β = −60.40, t = 4.81, p < .001). Thus, H8 is supported.

Figure 4. Moderation of social support (SS) in the link between voice behaviour (VB) and CWB

Discussion

The aim of our study was to develop and test the EL and normative conduct model through underlying mechanisms of SLT (Bandura, 1969, 1986), and COR theory (Hobfoll, 1989, 2002). Accordingly, we considered trait affect as an important mechanism, which was neglected earlier to be considered as a psychological process (Detert & Burris, 2007; Peng & Kim, 2020) in the relationship between EL and deviant normative conduct. Moreover, considering the underlying mechanism of SLT, we further considered the role of VB as a challenging employee behaviour at work (Van Dyne & LePine, 1998) in the relationship between EL and deviant normative conduct. The results of the study confirm all the relationships of the hypothesized model. The study findings show that there is a significantly positive relationship between EL and PA; however, the study reported that there is a significantly negative association between EL and NA. The findings further depict that PA enhances VB of subordinates, while NA depletes VB of subordinates. The subordinate VB significantly enhances OCB, and this relationship becomes stronger when a subordinate accepts social support from a manager. However, VB of subordinates leads to a significant decrease in CWB, and this relationship is further depleted when a subordinate accepts social support from manager.

Our study presents several contributions to theory and practice. Our study’s first crucial contribution to the theory is that we develop and test a theoretical model of EL and deviant normative conduct at work that uniquely synthesizes SLT and COR theories with critical psychological and contextual factors and conditions. There is enough evidence on the impact of EL impact on normatively (un)desirable conduct at work (Bedi et al., 2016; Ng & Feldman, 2015). However, with few exceptions (e.g., Peng & Kim, 2020; C. Yang et al., 2016), research on EL and employee’s deviant conduct at work have not reported empirical evidence. Moreover, these limited works have not reached a consensus on how EL impacts OCB and CWB simultaneously. The study at hand also provides empirical evidence about the EL differential influence on normatively (un)desirable conduct at work.

The second critical contribution is that our study adds to the literature studying and confirming the role of psychological process, i.e., trait affect (PA and NA) and VB simultaneously as serial mechanisms in the association between EL and deviant normative conduct, i.e., OCB and CWB. The study findings indicated that when the psychological processes (operationalized by trait affect) and challenging behaviour of subordinates (operationalized by VB) are tested as serial mechanisms, the EL and normative conduct at work (OCB and CWB) model has deviant explanatory power. These findings are in line with the recent call on leadership process theory development (Fischer et al., 2017; Peng & Kim, 2020; Van Knippenberg & Sitkin, 2013) and present new insights for future theory development on EL. This study empirically confirmed that the association between EL and OCB and CWB is not plain; it has crucial multiple psychosomatic and situational underlying mechanisms.

The third significant contribution of our study is the verification of social support as a crucial boundary condition in the association between subordinate VB and OCB and CWB. Social support was found to be a critical enhancer of the association between subordinate VB and OCB, while at the same time, social support is also identified as an essential suppressor of the association between subordinate VB and CWB. Our study, these findings are important and adds to the COR theory (Hobfoll, 1989, 2002), as our study reported the overlooked role social support role in the VB and deviant normative conduct relationships. Thus, these findings of this study enable future studies in findings a critical boundary condition that acts simultaneously as an enhancer and suppressor in the VB and deviant normative conduct relationships.

Our study’s first crucial contribution to the theory is that we uniquely add to the EL and normative conduct at work literature by suggesting the vital role of PA (positive trait) and NA (negative trait) simultaneously as psychological processes in explaining subordinate VB, which in turn has a deviant influence on normatively un(appropriate) conduct. In particular, these findings of our study are crucial and adds to knowledge by investigating a sample from non-individualistic culture, i.e., China, as leadership theorists asked for in-depth research on collectivist cultures (Wang et al., 2017).

Our study’s conceptual model of EL and deviant normative conduct at work presents several implications for managers. Our study’s first managerial implication is that supervisors and leaders should understand that if they want to boost OCB and to deplete the detrimental affect CWB of employees simultaneously, EL does’ matter. Our study’s second managerial implication is that a higher level of EL does matter as it increases PA and depletes PA simultaneously. Managers should understand that when subordinates comprehend leaders’ ethical behaviour, then their cognitive state is positive, (Chen & Hou, 2016; Hoyt et al., 2013), while, if subordinates perceive unethical behaviour, their cognitive state is inclined towards negativity (Ariely, 2012). Such divergent cognitive states of subordinates (cognitive state) can result in differential VB of subordinates, which in turn can lead to deviant normative conduct at work. Leaders and supervisors should better realize the ethical leadership approach, for instance, using EL behaviour to improve PA rather than NA to encourage subordinates to raise their voices to achieve OCB organizational goal instead of engaging in CWB. Our study’s third managerial implication is that leaders and supervisors should understand that subordinates have deviant emotional feelings, and they are usually in search of opportunities for advice and support from others (Morgeson & Humphrey, 2006). Leaders and supervisors should understand the social support role as an enhancer and suppressor; when a higher level of social support interacts with subordinates VB than it can enhance OCB and suppress CWB.

Even though the study at hand offers various triggering contributions to the theory and practice, yet few limitations are also reported. First, although we performed SEM analyses to check the hypothesized linkages simultaneously to establish causality, yet due to the employment of the cross-sectional data, the findings of the study should be generalized with care. Second, investigating a sample from non-individualistic culture, i.e., China, is one of the silent contributions of our study; still, care is needed in generalizing the study findings to Western culture. Accordingly, we recommend future studies to conduct a comparative study on the EL and normative conduct model. Third, our study is also limited to consider only trait affect as a psychological process. Future studies are recommended to use different psychological processes factors using different theoretical foundation will significantly add to the literature, such as future research can use the theory of trust (McAllister, 1995) as a crucial underlying mechanism (can be operationalized by trust, i.e., cognitive and affective trust; (J. Yang et al., 2009) to increase the explanatory power of this study hypothesized model. In addition, future studies can also infer causal linkages of the model through designing an experimental intervention study design, such as the design of EL behaviour climate to foster PA and NA relationships mainly to check and confirm whether the underlying processes caused increases or decreases by individual trait affect.

Conclusion

To conclude, this study uniquely synthesizes social learning and COR theories to EL and normatively (un)desirable conduct, i.e., OCB and CWB at work. This study developed and empirically validated a hypothesized model of EL, and OCB and CWB. This conceptual model also explained the role trait affect (PA and NA) as a psychological process and subordinate VB as a challenging behaviour in the relationship between EL and deviant normative conduct. Finally, the role of social support as a critical boundary condition is recognized in the relationship between subordinate VB and deviant normative conduct. This study used a sample from non-Western culture (i.e., Chinese), and therefore, the findings can be generalized to other collectivist cultures.

Acknowledgments

We would like to thank the Associate Editor Dr. Peter Caputi and two anonymous reviewers for their honest, helpful, and constructive comments.

Disclosure statement

The authors declare that they have no conflict of interest.

Additional information

Funding

The research was supported by the National Natural Science Foundation of China (72074024; 72041002) and by the Grant of Key Program for Beijing Educational Science Planning during the 13th Five-Year Plan Period, 2020 (Exploring the Influence of University Students’ Practical Activities on Leadership Emergence, No. CEAA2020047).