Ethical leadership is defined as “the demonstration of normatively appropriate conduct through personal actions and interpersonal relationships, and the promotion of such conduct to followers through two-way communication, reinforcement, and decision-making” (Brown et al., 2005). If a leader is to have a reputation for being ethical, he/she must be an ethical person and an ethical manager (Trevino et al., 2000; Ahn et al., 2018). As an ethical person, he/she has ethical characteristics such as honesty and integrity (Caldwell et al., 2008). As an ethical manager, he/she adopts ethical strategies that influence organizational ethics and behavior (Peng and Wei, 2020). Some studies on ethical leadership have confirmed that it has a positive effect on subordinate’s performance (Walumbwa et al., 2011), work effort (Haller, 2018), creativity (Feng et al., 2018; Mo et al., 2019), ethical behavior (Moore et al., 2019; Halbusi, 2021), such as prosocial behavior (Piccolo et al., 2010; Kacmar et al., 2011), green behavior (Saleem et al., 2020) and OCB (DeConinck, 2015; Gerpott et al., 2019).

The influence of ethical leadership on employees’ ethical behavior can be explained from different perspectives. From the social learning perspective, ethical leaders are role models and will be imitated by their employees (Bandura and Walters, 1977; Bandura, 1986). Ethical leaders are altruistically motivated, and they are likely to take actions against unethical behavior (Trevino et al., 2000; Brown et al., 2005; Marquardt et al., 2020). As a result, employees will look to ethical leaders as role models. From the social exchange perspective, employees choose their actions primarily based on their relationship with their leader (Walumbwa et al., 2011; Zhou, 2020). Ethical leaders who are caring, fair, and concerned about their employees can earn their trust and loyalty. In turn, employees are willing to reward their leaders by proactively offering constructive ideas and suggestions (Walumbwa and Schaubroeck, 2009). From the social norm perspective, ethical leaders create ethical behavior norms through communications and interactions among colleagues and rewards and punishments within the organization (Keck et al., 2020). Once moral or correct codes of conduct are established, they will promote employees’ ethical behavior just like voice behavior (Avey et al., 2012).

Voicing is “a discretionary communication of ideas, suggestions, concerns, or opinions about work-related issues with the intent to improve organizational functioning” (Burris et al., 2008; Detert et al., 2010). As an extra-role behavior, to voice is a bold choice made by employees for the organization’s interests (Morrison, 2011; Lee et al., 2017).

However, voicing is inherently challenging compared to other forms of extra-role behavior, such as helping behavior (Li and Tangirala, 2021). Voicing implies that employees point out flaws in organizational procedures and errors or mistakes of leaders or employees, which may imply irresponsibility on the part of leaders and dereliction of duty on the part of employees. Therefore, it may embarrass leaders or colleagues and damage their interpersonal relationships with vocal employees (LePine and Van Dyne, 1998; Hsiung, 2012; Hussain et al., 2019).

Considering these potentially challenging consequences, the antecedents of voice behavior are complex and involve a delicate balance of consequences in deciding whether or not to vocalize (Li and Tangirala, 2021). Morrison outlined two categories of factors that induce voice behavior: motivators and inhibitors. Motivators are the driving force behind voice behavior, while inhibitors are the restraining forces that pull employees toward silence and reduce their likelihood of speaking. Morrison notes that while many motivators have been empirically confirmed (including ethical leadership), other motivators factors have received less empirical attention, including affect (e.g., affective commitment).

In conclusion, ethical leadership can influence employees’ voice behavior through various mechanisms. However, previous empirical studies on the relationship between ethical leadership and employees’ voice behavior have only emphasized the affective effect of ethical leadership on employees and lacked attention to the affective impact of ethical leadership (Liu et al., 2017). As an affective factor, affective commitment can provide a new mechanism to explain voice behavior according to AET theory (Knoll and Redman, 2016). Affective commitment can lead to persistence during the action, even in the face of conflicting motivations or attitudes. It may even lead individuals to behave in ways that may be contrary to their self-interest from the perspective of a neutral observer (Herscovitch and Meyer, 2002). Thus, it is an essential complement to existing motivational theories and provides a new understanding of voice behavior in the work environment. One of the main objectives of this study is to extend the research on the relationship between ethical leadership and employee voice by introducing a new mediator, affective commitment.

This study tries to develop a conceptual model based on affective commitment and Affective Events Theory (AET) literature. Our first theoretical foundation is AET, which is mainly used to explain the relationship between the events experienced by organizational members in the workplace, affective reactions, and the attitudinal and behavioral changes caused by these affective reactions (Weiss and Cropanzano, 1996; Brief and Weiss, 2002). Affective reactions emphasize the individual’s psychological experience and have a multidimensional structure; they manifest as a short-term intense emotional experience or a relatively stable and lasting affective state (Weiss et al., 1999; Beal et al., 2013). This study focuses on this enduring affective state, which can influence an individual’s attitude and behavior toward work and organization (Judge et al., 2017). Our second theoretical foundation is literature on affective commitment.

On the one hand, affective commitment is defined as “the employees’ affective attachment to, identification with, and involvement in a particular organization, essentially representing the desire to stay in the organization” (Meyer and Allen, 1984; Meyer et al., 2004). According to Meyer and Allen (1991), affective commitment is a state of affective attachment that employees develop to an organization, a long-term, relatively moderated form of affective response. We believe that the series and accumulation of employees’ experience at work can change the long-term affective response. On the other hand, Herscovitch and Meyer’s extended three-components model indicated that employees’ affective commitment is associated with higher levels of behavioral support for change than other forms of commitment (Herscovitch and Meyer, 2002). Employees’ voice behavior symbolized a high degree of change orientation because voice represented employees’ questioning of the status quo and implied a tendency to change the status quo to solve problems. Based on this, we choose AET and the literature on affective commitment as the basis for our theory.

Employees’ traits may also influence their organizational commitment. Thus, in light of the influence of ethical leaders on employees’ voice behavior, personal traits of employees, such as moral disengagement, would affect this correlation. This study examines the moderating role of moral disengagement, a key personality trait associated with employees’ perception of ethical and unethical behavior.

Moral disengagement is defined as “an individual’s propensity to disengage morally — that is, an individual inclination to behave unethically without feeling guilt or distress” (Bandura, 1999; Newman et al., 2020). According to Bandura’s theory, moral disengagement helps reduce discomfort and self-blame when people act against their ethical standards (Gonzalez, 2018; Newman et al., 2020). Moral disengagement also provides reasons to rationalize unethical and unjust behavior without feeling upset (Fida et al., 2018; Lian et al., 2020). Thus, it plays a crucial role in explaining how individuals can engage in human atrocities such as political and military violence or corporate wrongdoing and corruption without apparent cognitive distress.

Bandura proposed that moral disengagement occurs through three cognitive mechanisms: relabeling unethical or unjust behavior as not unethical, alienating and distorting the harmful consequences of unethical or unjust behavior, and dehumanizing the victims of unethical behavior (Bandura, 1999, 2002). Moral disengagement is a process of rationalizing unethical behavior (Lian et al., 2020). It can also be seen as a personal characteristic that invalidates moral self-regulation and allows individuals to engage in unethical behavior without feeling guilty or uncomfortable (Moore, 2008; Cheng et al., 2019; Schaefer and Bouwmeester, 2020). This research viewed moral disengagement as an unchangeable individual characteristic (Detert et al., 2008). It assumes that moral disengagement moderates the relationship between ethical leadership and affective commitment and the relationship between affective commitment and voice behavior.

This study adopted a cross-sectional design with an individual level of measurement and analysis. Data were collected from only one organizational hierarchy. Specifically, we collected data from 15 retailing companies located along the southeastern coast of China. Retail is particularly close to people’s daily lives, and retailers need to gather their employees’ voices on customer feedback, product quality control, etc. Previous sound studies have collected samples from a single industry (e.g., Walumbwa and Schaubroeck’s research into a large financial institution) (Walumbwa and Schaubroeck, 2009) because this focus has a substantial advantage, that is, “unknown sources of variance due to the organization type can be controlled” (Near et al., 2004). We chose the retail industry as our research sample, following the same logic. The specific sampling process was as follows:

In terms of random sampling, we took a series of measures to ensure the randomness of the sample. As previous scholars asked for help from directors or supervisors (see Kacmar et al., 2013), our sampling process was also conducted with the help of HR. Firstly, we set a series of principles of random sampling designed for this study, and then asked HR to select 20 ∼ 60 employees randomly from different departments of each company according to this random sampling method; after that, we got a total of 600 employees as the sample, and we randomly assigned a three-digit code to each employee. Secondly, before filling in the formal questionnaire, researchers gave detailed instructions and explanations for employees, emphasizing the anonymity and confidentiality of the whole data collection process, and repeatedly stressed that participation was voluntary, so the voluntariness of the respondents also ensured the randomness of sampling; Thirdly, HR issued questionnaires to those selected employees (as some employees are away from the company on business or for other reasons, the number of questionnaires issued each time is about 540).

As mentioned before, this study adopted a three-wave data collection method to reduce the common method variance (CMV) (Podsakoff et al., 2003). Hansen et al. (2013), Liao et al. (2015) and Lee et al. (2017) chose a 1-month interval when studying the influence of ethical leadership to fully observe the impact of ethical leadership on outcome variables while reducing CMV. We also adopted a 1-month interval in the data collection process following their recommendation.

At Time 1, each envelope was marked with a three-digit number and distributed to employees. Envelopes were handed out to 543 employees because some did not show up for work for business reasons. We then received a total of 282 responses, including respondents’ demographic information (e.g., age, gender, education, position), department size, and items about ethical leadership. One month later, at Time 2, we collected the data on affective commitment and moral disengagement, and this time we received 278 responses. At Time 3, 1 month after Time 2, we collected the questionnaires focusing on voice behavior and got 275 responses in total. Finally, 239 matching questionnaires were identified by matching the three-wave questionnaires with three-digit numbers. We then omitted seven invalid questionnaires (where over 5% of the data were missing), leaving 232 valid samples.

Regarding sample size adequacy, Bentler and Chou (1987) suggested that the appropriate sample size is about 5 to 10 times the number of items to run the SEM. Generally, N = 100–150 is contemplated as the least possible sample size for conducting SEM (Ding et al., 1995; Kyriazos, 2018). Besides, we also calculated the sample size using software, the first one was G*Power 3.1, α = 0.05, power = 0.95, f² = 0.15, the calculated sample size was 89 cases; the second one was PASS15, α = 0.05, power = 0.95, f² = 0.15 (the medium effect size), the calculated sample size was 97 cases.

To sum up, the sample size used in this study (N = 232) was much larger than those calculated by the software, and at the same time fulfilled standards recommended by previous scholars (Bentler and Chou, 1987; Kock and Hadaya, 2016; Kyriazos, 2018), then the sample size of this study seemed adequate and justified to run the SEM for data analysis, fulfilling the minimum sample size requirement.

Table 1 provides the sample characteristics, 40.5% were male, 59.5% were female, 41.4% were 31–40 years old, and 69.0% had a bachelor’s degree, 58.2% of the sample held low-level positions. As for tenure, 35.3% had been in their jobs for more than 8 years. Only 0.9% of the samples belonged to the R&D department, while 50.9% belonged to the management department. Finally, nearly half have more than 20 employees.

Since all the multi-item measures in this study were initially constructed in English, we developed Chinese versions for all the measures following the commonly used back-to-back translation procedure. After that, we pre-tested four key variables and assessed all of these items on a 5-point-Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree).

Although the concept (leadership) seems to be multileveled by nature, researchers are interested in individuals’ perceptions of ethical leadership in the current study (Cheng et al., 2019). We examine the measurement and structural model simultaneously with Partial Least Squares Structural Equation Modeling (PLS-SEM) using SmartPLS 3.0 (Ringle et al., 2015). PLS-SEM enables users to create a single theoretical model and simultaneously examine the direct, indirect, and moderating effects of ethical leadership on affective commitment and voice behavior. A confirmatory composite analysis is the appropriate two-step method for PLS-SEM analysis (Hair et al., 2020). The first step is an evaluation of the measurement model, and the second step is an evaluation of the structural model (Sarstedt and Cheah, 2019). Below we detail the data analysis and results of this study.

The descriptive statistics, including means, standard deviations, and correlations for all the variables included in the conceptual model, can be found in Table 2. These results indicate multiple significant relationships between constructs in the structural model, and many instances are consistent with previous findings in the literature. The results also support the reliability and validity of the theoretical measurement models.

Measurement model assessment begins with evaluation and confirmation of reliability and validity of the outer measurement models. Results of the initial measurement model assessments are shown in Table 3. All indicators exhibited acceptable item reliabilities. All measurement models (constructs) were above the recommended levels of.70 for composite reliability. Overall, all requirements for measurement model reliability were well above recommended minimum guidelines (Hair et al., 2021). The convergent validity of the measurement model can be assessed by the Average Variance Extracted (AVE). The AVEs of most constructs were well above the minimum recommended level of.50 (Hair et al., 2021), while the AVE of Moral disengagement was 0.408, which was not qualified. Since Henseler et al. (2015) proposed that Fornell and Larcker’s (1981) criterion and cross-loadings were not sufficient for discriminant validity, we then assessed the discriminant validity between constructs using the heterotrait-monotrait (HTMT) method. The HTMT values were between 0.194–0.468, less than the strict standard of 0.85 (Kline, 2015). Thus, discriminant validity among all constructs was confirmed.

The second step of the confirmatory composite analysis process is the structural model’s assessment (Hair et al., 2020). We used PLS Algorithm to analyze the structural model. Standardized Root Mean Square Residual (SRMR) could be used to evaluate model fitness. When the SRMR of the Saturated Model and the Estimated Model is less than 0.08, they have a good fit (Hu and Bentler, 1999), and the smaller the gap between the values of the Saturated Model and the Estimated Model, the better (Henseler et al., 2015). The results of this structural model showed that the values of the saturation model and estimation model were 0.061 and 0.063, respectively, so the model fit well.

Then, we examined the path coefficients and significance levels for the hypothesized relationships. These metrics were obtained by executing the PLS bootstrapping procedure. We used 5,000 samples to produce bias-corrected confidence intervals for each coefficient for this procedure. The hypothesized direct relationships were examined first, and then the hypothesized indirect relationships. Figure 2 provides an overview of the results.

Our findings contribute to the literature on ethical leadership and voice behavior. First, this study extends the research on ethical leadership by adding a substantive mediator to explain the influence of ethical leaders on employee voice behavior. Second, our study provides a new perspective that employees’ affective commitment to the organization is a powerful motivation for employees to speak up when motivated by ethical leadership. When employees have this kind of attachment to the organization, they are willing to accept the risks involved (e.g., extra work and cost) and challenge the status quo by expressing their opinions. Although some previous research has described how ethical leadership promotes employee voice behavior, most focus on rational mechanisms. Employees choose whether or how to voice based on the benefits and costs of voice behavior and the success or failure of voice behavior. The existing mediators, such as psychological safety (Walumbwa and Schaubroeck, 2009; Zhou and Wang, 2018), organizational ethical culture and climate (Demirtas and Akdogan, 2015; Bai et al., 2019), unethical climate (e.g., political climate) (Cheng et al., 2019), and individuals’ moral efficacy (Lee et al., 2017), are all understood from the rational perspective. These mediators seldom consider the complex role of leadership on outcomes. Therefore, based on AET, this study proposes that affective commitment to the organization serves as the mediating mechanism by which ethical leadership stimulates employees’ voices behavior, enriching the understanding of the relationship between ethical leadership and unique extra-role behavior.

Second, our findings suggest that the crucial role of employees’ moral disengagement as a condition that strengthens or weakens the expected response to perceived ethical leadership. Previous research has not considered the personal traits of subordinates, which may be related to the leader’s influence and the degree to which the leader assimilates them. Employees with a low degree of moral disengagement are more likely to experience psychological comfort or empathy when observing or perceiving the ethical behavior of leaders. Therefore, for employees with a low degree of moral disengagement, leaders’ ethical behavior has a more significant impact on their affective commitment to the organization. On the contrary, when the moral disengagement degree of employees is high, it is difficult for ethical leadership to promote their affective commitment to the organization.

Third, our study enriches proactive behavior theory (including voice). Many previous studies believe that organizational commitment promotes employees’ proactive behavior (Lapointe and Vandenberghe, 2018), while others do not. For example, Dean and Greene (2017) suggest that attachment to an organization blocks voice (Dean and Greene, 2017). In research by Parker et al. (2006), empirical evidence also indicates that affective organizational commitment has nothing to do with proactive behavior (an integrating measure including proactive idea implementation and proactive problem solving). Our results show that the shift from affective commitment to extra-role behavior such as voice depends on individual traits, especially moral disengagement. While affective commitment is associated with an employees’ willingness to contribute to the organization, it does not necessarily bring attention to issues or motivate them to take action, especially by voice. Voice behavior is closely related to an individual’s judgment of right and wrong and internal moral standards. It involves the assessment of problems existing in the organization and the willingness of employees to take responsibility for their suggestions. Therefore, an individual’s moral disengagement can explain what conditions affective commitment will bring voice behavior. Compared with just saying (e.g., voice behavior), doing (e.g., proactive behavior) needs to overcome more difficulties and require the actor to have a higher moral standard. Further research should consider moral disengagement as a moderator to explain the relationship between the attachment to the organization and more holistic, proactive behavior.

Finally, the overall integrated moderated mediation model provides substantial evidence that the extent to which affective commitment mediates the relationship between ethical leadership and voice depends on the degree of an individual’s moral disengagement. Although previous studies have confirmed that organizational ethical environment or personal factors are related to the relationship between ethical leadership and voice (Zhu et al., 2015; Bai et al., 2019), they remain silent about the conditions under which the mediating effect of situational or personal factors is amplified or attenuated. By identifying this boundary condition, this study contributes to a more accurate understanding of the role of ethical leadership in organizations. Also, it helps explain “the complex ethical leadership-employee performance relationship” (Walumbwa and Schaubroeck, 2009).

Our findings may have the following contributions which can be used in organizations in the future. First, our study proved the positive effect of ethical leadership on employees’ psychology and behavior. Many organizations choose managers more based on their competence and performance in the modern business society. However, they do not pay enough attention to the ethical or moral nature of the manager. Our research demonstrates that ethical leadership encourages employees to speak up, which is vital for business development and increasingly important in the VUCA (volatility, uncertainty, complexity, and ambiguity) environment (Detert and Burris, 2007). In a VUCA environment, it is increasingly difficult for its leadership to grasp market trends and consumer behavior trends. To catch these market changes, organizations need to rely on all employees’ knowledge, wisdom, and information and their advice, so ethical leadership has a more critical role in VUCA environments (Schoemaker et al., 2018). Therefore, when selecting and promoting managers, companies should check leaders’ morals and ethics.

Second, our findings indicate that employees’ affective commitment to the organization promotes employees’ voice behavior. In the modern business world, the relationship between companies and employees is very fragile, and companies do not pay much attention to cultivating feelings with employees (Eisenberger et al., 2002). Companies and employees have more transactional relationships than commitment-based relationships. When companies face strategic transformation and business failure, they will ruthlessly lay off employees so that employees will not do their best for the company and give the advice to contribute to its development (Hom et al., 2017). Therefore, companies should care about their employees, respect them, provide competitive salaries and opportunities to enhance their employees’ commitment to the organization.

Third, our results indicate that moral disengagement is vital in promoting employees’ voice behavior in ethical leadership’s function. Employees with a high sense of moral disengagement are unresponsive to ethical leadership and organizational problems. They are reluctant to contribute to the development of the organization. Therefore, organizations should also pay attention to the moral tendencies of their employees when selecting them. They should not choose employees who can easily justify their misconducts; instead, they should select employees who can frequently reflect on themselves and correct their wrongdoing carefully. At the same time, the organization should continually communicate with employees about ethics to increase their sensitivity to ethical issues and guide them to distinguish right from wrong in organizational phenomena more accurately.

Despite these contributions, our research has some limitations. First, data analysis was based on a sample drawn from the Chinese retail industry, limiting our findings’ generalizability. Therefore, future research can replicate this study in other countries with different cultures or contexts to examine and enhance the validity and generalizability.

Second, this study collected data from a single set of respondents. We used perceptual data of the respondents. Although objective and subjective evaluations are not the same, perception measurement is generally an accepted way to measure leadership and employee behavior (Lee et al., 2017; Cheng et al., 2019). Although we collected data in three phases and used statistical tools to examine common method bias, we also used the respondents’ anonymity to ensure the social desirability bias was minimized; we could not completely rule out the possibility of this form of bias. Using alternative measurement methods (e.g., using data from multiple sources and examining both other- and self-reported sources) may help reduce potential perceptual bias in ratings. Future research could use a multilevel approach to analyze the effects of ethical leadership.

Third, regarding the control variables at different times (e.g., control for t1 and t3 when using a variable measured at t2), we used the same control variables to seek uniformity, making it difficult to measure the effects of different control on specific variables. Different control variables should be used in future studies to control their influence on the outcome variables better.

Fourth, while we did examine an integrated theory-related moderated mediation model, other mechanisms could help explain the relationship between ethical leadership and employee voice. Future research should examine different mediating factors such as moral identity in more detail and explore the moderating effects of other personality traits such as emotional intelligence, political skills, situation-specific leverage, the locus of control.