By enacting voice—that is, speaking up about ideas, concerns, information, or opinions about work-related issues (Van Dyne, Ang, & Botero, 2003)—employees provide critical information that can help organizations make informed decisions, thereby improving performance (see Bashshur & Oc, 2015, for a review). It is thus important for managers to understand the conditions that facilitate employee voice (Morrison, 2011). Job satisfaction is the most widely examined antecedent of voice behavior (Chamberlin, Newton, & Lepine, 2016). Informed by social exchange theory (Blau, 1964) and the reciprocity norm (Gouldner, 1960), voice scholars have argued that more satisfied employees believe that they have received more benefits from their organizations and thus feel obligated to contribute more ideas (e.g., Morrison, Wheeler-Smith, & Kamdar, 2011; Van Dyne & LePine, 1998). One the other hand, scholars have also shown that less satisfied employees perceiving more problems at work are more willing to change the status quo through voice so as to regain satisfaction (Farrell, 1983; Hirschman, 1970; Tangirala & Ramanujam, 2008; Zhou & George, 2001). We integrate these two distinct but complementary perspectives by exploring a curvilinear relationship between job satisfaction and employee voice. By doing so, we hope to contribute a more complete and accurate explanation for this important relationship.

Employees are satisfied with their jobs to the extent that what they want from their jobs are met (Locke, 1969). At low levels of satisfaction, jobs do not live up to employees’ standards, which makes the problems at work salient to them. In this case, they may be motivated to engage in voice to change the status quo so as to regain satisfaction. Conversely, when job satisfaction reaches a certain level, at which their desires are fulfilled by their organizations, employees enter a state of gains and begin thinking about reciprocating to their organizations in terms of voicing suggestions for improvement. In other words, the motivation to address problems that are unsatisfying is stronger at lower satisfaction and the motivation to reciprocate to the organization is stronger at higher satisfaction. Therefore, we argue that there is a U-shaped relationship between job satisfaction and employee voice. In addition, the extent to which individuals view voice as a way to help organizations rather than as a way to help themselves to regain satisfaction is likely to strengthen the positive relationship between job satisfaction and voice at high levels of satisfaction and weaken the negative relationship between them at low levels of satisfaction (i.e., the U curve tilts to the left). Thus, we argue that a prosocial normative voice belief (hereinafter, prosocial voice belief), which refers to a normative belief that employees should help their organizations through voice behavior, moderates the relationship between job satisfaction and voice.

Employees who speak up incur personal costs because they may be punished for challenging the status quo and creating troubles for others (Burris, 2012; Morrison, 2011; Van Dyne, Cummings, & Parks, 1995). The need to reciprocate or to regain satisfaction must outweigh the costs for employees to really take action to engage in voice. A self-protective implicit voice belief (hereinafter, self-protective voice belief) captures a socially acquired and taken-for-granted belief about when and why voice is risky and inappropriate (Detert & Edmonson, 2011). When this belief is strong, employees perceive voice behavior as costly and thus less attractive as a way to reciprocate or to address problems. Therefore, we expect self-protective voice belief to attenuate the relationship between job satisfaction and voice at both low and high levels of job satisfaction. In other words, the relationship between job satisfaction and voice may become less U shaped when employees have a stronger self-protective voice belief.

In summary, we integrate two motivations of voice behavior into the curvilinear relationship between job satisfaction and voice and study the moderating effects of two voice beliefs to examine the motivational processes underlying the curvilinear relationship. By doing so, we aim to contribute to the voice literature in several ways. Theoretical precision may be hindered by the automatic assumption of linearity (Edwards & Berry, 2010). By modeling a more complex nonlinear relationship, our study reveals that employees tend to speak up not only when they are highly satisfied, such that they want to reciprocate their organization’s kindness with constructive ideas, but also when they are highly unsatisfied, such that they need to address the problems to improve their condition. This more accurate understanding of the relationship between job satisfaction and voice can better inform managers about which employees are likely sources of voice and how to motivate their voice behavior.

As Morrison (2014, p.184) noted in her review of voice research, “Voice is primarily prosocial, but this does not mean that it lacks benefits for the actor, nor that the actor will fail to consider those benefits. It is therefore reasonable to assume that when employees are deciding whether to engage in voice, they may consider not just how this behavior could lead to organizational or unit-level improvement, but also how it could potentially advance their own interests.” Our study demonstrates how these different voice motives may come to play together. Specifically, we argue that the relative salience of the voicer’s motives to help themselves and to help the organization varies across different levels of job satisfaction – that is, employees are more likely to think about helping their organization after they have solved their own problems. Our study also shows that two individual beliefs (i.e., prosocial and self-protective) about voice behavior influence the extent to which the two voice motives take effect.

Finally, voice research has focused on the influence of organizational context, supervisor behavior, and employee work attitudes and dispositions (see Morrison, 2011, for a review). By and large, the literature has assumed that people have similar beliefs about and attitudes toward voice behavior. As an exception to this assumption, Detert and Edmonson (2011) conceptualized self-protective voice belief as representing an individual’s belief about the risk and cost of voice behavior and found that this belief has a significant influence on employee silence. Our study moves beyond examining the main effect of this indiviudal belief and suggests that it also moderates how factors such as job satisfaction affects voice. We also contribute a new individual belief about voice behavior, the prosocial voice belief, and examine its moderating effect on the relationship between satisfaction and voice. These nuanced insights shed greater light on when employees speak up in organizations.

Hypotheses development

Job satisfaction and voice behavior

Employees are satisfied with their jobs to the extent that job features such as the work, pay, supervisors, coworkers, and developmental opportunities fulfill their desires (Kendall, Smith, Hulin, & Locke, 1963). There are two theoretical perspectives on how employees respond to different levels of job satisfaction when making voice decisions. The first is based on social exchange theory (Blau, 1964) and highlights voice as employees’ reciprocation. Specifically, employees who are more satisfied with their jobs believe that they have received more from their organization; as such, the reciprocity norm (Gouldner, 1960) obligates them to pay back their organization by engaging more in behavior that benefits the organization, such as voice (e.g., Liang, Farh, & Farh, 2012; Morrison et al., 2011). The second perspective emphasizes that voice can help employees regain job satisfaction. Although voice is a citizenship behavior that can benefit an organization, this behavior may also be instrumental to the employees’ personal benefit (Bolino, 1999; Grant & Mayer, 2009; Rioux & Penner, 2001). From this perspective, less satisfied employees have a tendency to voice more because they not only perceive more issues that need to be changed but also benefit more from making the change (Farrell, 1983; Hirschman, 1970; Tangirala & Ramanujam, 2008; Wu, Tang, Dong, & Liu, 2015; Zhou & George, 2001). For example, a constructive suggestion may both improve organizational functioning and ease the speaker’s job (Morrison, 2014). Employees who engage in voice have an opportunity to promote the issues they are concerned about and direct social attention and resources to these issues (Ashford, Rothbard, Piderit, & Dutton, 1998). Voice can also serve as a means to gain personal control over one’s job (Tangirala & Ramanujam, 2008), affirm a sense of self in the workplace (Ashford & Barton, 2007), and enhance social image and influence at work (Fuller, Barnett, Hester, Relyea, & Frey, 2007). These benefits can help employees regain their job satisfaction.

The two theoretical perspectives characterize two distinct voice motivations. We argue that as job satisfaction increases, employees’ mindset shifts from fixing their losses to reciprocating their gains. The motivation to address problems so as to regain job satisfaction is stronger at low levels of job satisfaction—before employees’ standards for their jobs are fulfilled. In this state of loss, employees’ primary goal is to fix their problems, and reciprocation is less relevant. By contrast, the motivation to reciprocate is stronger at high levels of job satisfaction, at which organizations provide employees more than expected and problems are no longer salient to the employees. Neither motivation is strong at intermediate levels of job satisfaction, in which employees’ standards are just being met. These arguments suggest that employees speak up more at low and high levels of job satisfaction than at intermediate levels of job satisfaction. Thus, we propose the following:

  • Hypothesis 1. The relationship between job satisfaction and voice behavior is U shaped.

To further illustrate the motivation process underlying the U-shaped relationship between job satisfaction and voice behavior, we introduce two personal voice beliefs as moderators — prosocial voice belief and self-protective voice belief.

The moderating effect of prosocial voice belief

We argue that in employees’ voice decision, a prosocial voice belief strengthens the dominance of organizational concerns at high levels of job satisfaction and weakens the dominance of personal concerns at low levels of job satisfaction. A prosocial voice belief is an individual belief that employees should help their organizations through voice behavior. It is a particular type of personal norm that represents individuals’ expectations of their own behavior (Fishben, 1967). Although such normative beliefs may derive from social expectations, they are enforced through a self-based process—that is, conformity to these personal normative beliefs leads to self-approval, whereas violation results in self-deprecation (Deutsch & Gerard, 1955; Ehrhart & Naumann, 2004; Grant, 2012; Schwartz, 1973). People have a fundamental need to enhance their self-esteem (James, 1890; Rogers, 1959), which makes them conform to rather than violate personal normative beliefs (Schwartz, 1977). As individuals’ norms and goals direct their attention (Dijksterhuis & Aarts, 2010), a personal norm of helping organizations through voice is likely to direct employees’ attention from the value of voice behavior for themselves to the value of voice for the organizations.

When employees have a strong prosocial voice belief, they expect their voice to benefit the organization rather than themselves. Thus, they are more likely to view voice as reciprocation to the organization rather than as a way to improve their personal condition. For these people, the positive relationship between job satisfaction and voice is strengthened at high levels of job satisfaction, as their prosocial normative belief reinforces them to speak up for the benefit of the organization under this condition, and their negative relationship is weakened at low levels of job satisfaction, as speaking up for personal benefit under this condition contradicts their prosocial normative belief and may result in self-criticism and self-concept-related distress (e.g., Cialdini, Reno, & Kallgren, 1990; Garst, Harris, Kerr, & Lewandowski, 1997; Schwartz, 1973). When employees’ prosocial voice belief is weak, they are less likely to use voice to reciprocate the benevolence of their organization and more likely to take advantage of their voice to gain personal benefits, because such behavior does not violate their personal norm. For them, the relationship between job satisfaction and voice is more negative at low levels of job satisfaction and less positive at high levels of job satisfaction. In summary, we argue that a prosocial voice belief increases the dominance of organizational concerns in voice decisions (thereby strengthening the positive relationship between job satisfaction and voice) at high levels of job satisfaction and reduces the dominance of personal concerns (thereby weakening the negative relationship between job satisfaction and voice) at low levels of job satisfaction. Therefore, we hypothesize the following:

  • Hypothesis 2. Prosocial voice belief moderates the relationship between job satisfaction and voice, such that the positive relationship at high levels of job satisfaction is stronger and the negative relationship at low levels of job satisfaction is weaker for employees with a stronger prosocial voice belief.

The moderating effect of self-protective voice belief

Voice is a deliberate decision in which employees consider not only the benefits but also the costs of voice behavior (Ashford et al., 1998; Detert & Burris, 2007; Detert & Edmonson, 2011; Morrison et al., 2011; Ng & Feldman, 2012). One important cost of voice behavior is the potential damage to the speaker’s relationship with peers and supervisors (Van Dyne & LePine, 1998). In the eyes of conservative colleagues, employees who speak up may appear pushy or overly ambitious. Their suggestions may disrupt the performance routine and sometimes threaten other employees’ resources, which can strain relationships and cause conflicts (Bolino, Valcea, & Harvey, 2010). Voicers may also offend managers or other colleagues by pointing out their failures (Detert & Edmonson, 2011; Liang et al., 2012). Moreover, as a proactive extra-role behavior, voice often consumes a wealth of resources, such as time, mental energy, and even physical energy (Bolino et al., 2010), and thus can lead to role overload, job stress, and work–family conflict (Bolino & Turnley, 2005). We argue that when employees believe that voice behavior is costly, the influence of benefit considerations on voice decisions becomes weaker.

To understand the extent to which individuals believe voice is costly and risky, Detert and Edmonson (2011) proposed five self-protective implicit voice beliefs: a) making suggestions will be taken as personal criticisms or challenges to the person in charge; b) it is not safe to speak up if one does not have solid data, well-proved solutions, and firm answers to all possible questions; c) speaking up can be considered disloyal because it exposes mistakes of one’s boss or superiors; d) bosses in general dislike hearing challenges in public without private notice in advance; and e) voice will lead to retaliation and, consequently, to a stalled career. These self-protective beliefs orient individuals toward the potential negative consequences of voice behavior, such as loss of positive relationships with supervisors and colleagues, reputation damage, and adverse career consequences. These beliefs are considered implicit (or common sense) because they are developed and relied on subconsciously (Hong, Chiu, Dweck, Lin, & Wan, 1999). However, as stable and enduring individual beliefs (Epitropaki & Martin, 2004), these implicit beliefs have strong influences on individual judgments and reactions (Chiu, Hong, & Dweck, 1997; Dweck, Chiu, & Hong, 1995; Hong et al., 1999).

People instinctively avoid negative experiences and harm (Taylor, 1991). When employees have a strong self-protective voice belief, negative cues associated with voice behavior may stand out and be disproportionately weighted. To these employees, voice behavior is risky, so it is not worthwhile to obtain benefits through voice, whether for themselves or for their organizations (Detert & Edmonson, 2011; Wyer, 2004). Therefore, we expect self-protective voice belief to attenuate the relationship between job satisfaction and voice at both high and low levels of job satisfaction. In other words, the relationship is less U shaped for employees with a strong self-protective voice belief. By contrast, employees with a weak self-protective voice belief are less likely to equate voice behavior with adverse consequences (Detert & Edmonson, 2011). To them, voice is a more attractive way to solve problems that are unsatisfying or to reciprocate organizational support. Therefore, the U-shaped relationship between job satisfaction and voice is more likely to occur among employees with a weak self-protective voice belief. We hypothesize the following:

  • Hypothesis 3. Self-protective voice belief moderates the relationship between job satisfaction and voice behavior, such that the relationship is more U shaped for employees with a weaker self-protective voice belief.

Method

Respondents and procedures

While we believe the curvilinear relationship between job satisfaction and employee voice is generic, to demonstrate the importance of this effect, we selected a research context that allowed us to conduct a conservative test of the effect (Prentice & Miller, 1992.) Specifically, if we can find the curvilinear effect of job satisfaction on voice even in a context where voice is relatively difficult to influence, we are confident that the effect is indeed salient. As argued above, the relationship between job satisfaction and voice is less U shaped when voice is more costly. Speaking up is more costly in higher power distance cultures because this behavior constitutes a challenge to the managerial authority. Therefore, we tested our hypotheses on a sample recruited in China, which scores high on power distance.Footnote 1 Chinese leaders are characterized by a directive, hierarchical and control approach to leadership (Chen, Li, & Leung, 2017). Under such leadership, subordinates tend to perceive voice behavior as a risk and thus be reluctant to engage in such behavior.

We conducted a two-wave study of new graduates of a university in a major city in southern China. To increase the response rate, we issued an invitation to prospective respondents approximately 1 month before they graduated. We approached them in their classes and in the student center and obtained e-mail and mailing addresses from 350 students. We conducted the first survey approximately 6 months after the students graduated, as we expected that by then most would have found a job and begun working. We e-mailed the alumni a link to the online survey and mailed them paper copies of the questionnaire. Two hundred twenty alumni responded, yielding a response rate of 62.9%. Attrition occurred for various reasons, such as unwillingness to participate in the survey, loss of contact, unemployment, and continued study at a higher level. As our study targeted employees, we excluded alumni who had not yet found a job or were pursuing a higher degree. We applied the same procedure to the second survey conducted 2 months later. Of the 220 respondents to the first survey, 199 responded to the second survey, yielding a response rate of 90.5%. These individuals received monetary compensation for completing both surveys.

The final sample consisted of 199 new employees, 26% of whom were men. Respondents’ ages ranged from 21 to 29 years, and job tenure was between one and 6 months at Time 1. The respondents worked in a wide variety of industries: education and culture (18%), gambling (16%), financial services (16%), professional services (9%), other services (8%), the public sector (10%), travel (6%), trade (3%), information technology (3%), logistics (3%), manufacturing (2%), real estate (1%), insurance (1%), health care (1%), and others (7%). All respondents’ highest educational level was a bachelor’s degree.

Measures

We measured job satisfaction and prosocial voice norm and collected demographic information in the first wave and self-protective voice belief and voice behavior in the second wave. Six-point Likert scales with anchors ranging from “strongly disagree” to “strongly agree” were used in measures of all of the substantive variables.

Job satisfaction

We used the three-item scale developed by Cammann, Fichman, Jenkins, and Klesh (1983) to measure job satisfaction. A sample item is “All in all, I am satisfied with my job.” The reliability for this scale was satisfactory at .87.

Voice behavior

We measured voice behavior using the six-item scale developed by Van Dyne and LePine (1998). Sample items include “I develop and make recommendations concerning issues that affect this work group” and “I speak up and encourage others in this department to get involved in issues that affect the group.” Cronbach’s alpha of this scale was .84.

Prosocial voice belief

We operationalized this normative belief in terms of whether respondents believed that an employee should help his or her organization by engaging in voice behavior. We adapted Burris, Detert, and Chiaburu’s (2008) three voice items to measure the prosocial voice belief. Specifically, the respondents were asked to indicate how much they agreed or disagreed that they should help their organization by (1) informing managers about problems in their organization or department, (2) giving managers suggestions about how to make the organization or department better, and (3) speaking up to managers with suggestions for addressing employees’ needs and concerns. Cronbach’s alpha for this scale was .89.

Self-protective voice belief

We assessed self-protective voice belief using Detert and Edmonson’s (2011) 20-item scale. Sample items include “Someone who helps create a process or routine is likely to be offended when others suggest changes” and “Pointing out problems, errors, or inefficiencies might very well result in lowered job evaluations.” Cronbach’s alpha for this scale was .89.

Analytical approach

We followed Aiken and West’s (1991) recommended procedures to test our hypotheses. We mean-centered the independent variable (i.e., job satisfaction) and the moderators (i.e., prosocial and self-protective voice beliefs) before creating the squared terms and interaction terms. Our analysis for testing the curvilinear relationship between job satisfaction and voice is specified in the following equation:

$$ \mathrm{Voice}= b0+ b1\times \mathrm{job}\ \mathrm{satisfaction}+ b2\times \mathrm{job}\ \mathrm{satisfaction}\ \mathrm{squared}+\mathrm{error}\ \mathrm{term}. $$

To support Hypothesis 1, b2 needs to be positive and significant.

As we proposed differential moderating effects of prosocial voice belief and self-protective voice belief, the methods for testing these effects also differed. Hypothesis 2 states that a prosocial voice belief affects the slope (rather than curvilinearity) of the relationship between job satisfaction and voice. To test this hypothesis, we used the method Aiken and West (1991, p. 69) and Tangirala and Ramanujam (2008) recommended and examined the following equation:

$$ \mathrm{Voice}= b0+ b1\times \mathrm{job}\ \mathrm{satisfaction}+ b2\times \mathrm{job}\ \mathrm{satisfaction}\ \mathrm{squared}+ b3\times \mathrm{prosocial}\ \mathrm{voice}\ \mathrm{belief}+ b4\times \mathrm{job}\ \mathrm{satisfaction}\times \mathrm{prosocial}\ \mathrm{voice}\ \mathrm{belief}+\mathrm{error}\ \mathrm{term}. $$

To support the moderating effect of prosocial voice belief, b4 must be positive and significant. That is, at different levels of prosocial voice belief, the relationship between job satisfaction and voice has similar curves but different slopes at certain points of job satisfaction.

Hypothesis 3 states that self-protective voice belief affects the curvilinearity (rather than slope) of the relationship between job satisfaction and voice. We used another equation from Aiken and West (1991, p. 69-70) to test this hypothesis:

$$ \mathrm{Voice}= b0+ b1\times \mathrm{job}\ \mathrm{satisfaction}+ b2\times \mathrm{job}\ \mathrm{satisfaction}\ \mathrm{squared}+ b3\times \mathrm{self}-\mathrm{protective}\ \mathrm{voice}\ \mathrm{belief}+ b4\times \mathrm{job}\ \mathrm{satisfaction}\times \mathrm{self}-\mathrm{protective}\ \mathrm{voice}\ \mathrm{belief}+ b5\times \mathrm{job}\ \mathrm{satisfaction}\ \mathrm{squared}\times \mathrm{self}-\mathrm{protective}\ \mathrm{voice}\ \mathrm{belief}+\mathrm{error}\ \mathrm{term}. $$

To support the moderating effect of self-protective voice belief as proposed in Hypothesis 3, b5 must be negative and significant. That is, the quadratic relationship between job satisfaction and voice varies as a function of self-protective voice belief. The simple effect test must show that the relationship is more U shaped when self-protective voice belief is weaker.

Results

Before testing our hypotheses, we conducted a confirmatory factor analysis (CFA) to evaluate the discriminant validity of the focal variables in this study (i.e., job satisfaction, voice, prosocial voice belief, and self-protective voice belief). As the scale of self-protective voice belief has a large number of items and our sample size is not large, we used the five dimensions of self-protective voice belief as its indicators to reduce the number of parameters to estimate so as to enhance the estimation accuracy (Beauducel & Wittmann, 2005). As shown in Table 1, the four-factor model fit the data well: χ2 = 206.31, df = 113, comparative fit index (CFI) = .94, Tucker–Lewis index (TLI) = .94, root-mean-square error of approximation (RMSEA) = .065. All of the indicators were significantly loaded on their intended factors. The correlations of the latent constructs were low to moderate. The 4-factor model fitted the data better than its alternative models. In particular, the one-factor model yielded a poor fit: χ2 = 1053.25, df = 119, CFI = .36, TLI = .37, RMSEA = .199. These results suggested that job satisfaction, voice, prosocial voice belief, and self-protective voice belief were distinct and the common method bias is not a big concern in our data.

Table 1 Results of confirmatory factor analysis
Full size table

Table 2 displays the Cronbach’s alphas, means, standard deviations, and correlations among the variables examined in this study. We did not control for employees’ demographic characteristics (gender, age, and tenure) because they were not related to the dependent variable (voice) and because whether or not controlling for these characteristics did not change the results of our hypotheses testing (Carlson and Wu (2012).

Table 2 Means, standard deviations, and intercorrelations among variables
Full size table

Table 3 reports the results of the U-shaped curvilinear relationship between job satisfaction and voice and the moderating effects of prosocial voice belief and self-protective voice belief. We first regressed voice on job satisfaction and found that job satisfaction was positively related to voice (b = .24, SE = .05, p < .001). When we included the squared job satisfaction term in the model, we found that it had a positive and significant relationship to voice (b = .14, SE = .04, p < .01), and the R-square change was also significant (ΔR2 = .05, p < .01). This provides support for Hypothesis 1 about the U-shaped relationship between job satisfaction and voice behavior. The U-shaped curve reached its minimum point when mean-centered job satisfaction was −1.11 (i.e., the raw score of job satisfaction was 3.04). Figure 1 depicts this relationship. Following the suggestions of Preacher, Rucker, and Hayes (2007) and the procedures of Lam, Xu, and Loi (2018) and Kim, Lin, and Kim (Kim, Lin, & Kim, In press), we also used the Johnson–Neyman technique (Johnson & Neyman, 1936) to explore the values of job satisfaction for which the simple slope of the curvilinear relationship is significantly positive or negative. As Table 4 shows, when the value of job satisfaction was equal to or greater than 3.48 (b = .12, SE = .06, p = .05), the relationship between job satisfaction and voice behavior was positive and significant, and the slope increased as the value of job satisfaction increased. Conversely, when the value of job satisfaction decreased to less than 3.04, the slope became negative. When job satisfaction decreased to 2, the negative slope became marginally significant, and when it was 1.5 or less (b = −.43, SE = .22, p = .05), the negative slope was significant.

Table 3 Results for the U-shaped relationship and the moderating role of prosocial and self-protective voice beliefs
Full size table
Fig. 1
figure 1

The U-shaped curvilinear relationship between job satisfaction and voice behavior

Full size image
Table 4 Simple slopes of the relationship between job satisfaction and voice behavior for a range of job satisfaction values
Full size table

We predicted that prosocial voice belief would moderate the slope of the relationship between job satisfaction and voice behavior (Hypothesis 2). We entered prosocial voice belief and its interaction with job satisfaction into the equation. The results (see Model 3 in Table 3) showed that prosocial voice belief was significantly related to voice (b = .28, SE = .07, p < .001), and the coefficient of the interaction between job satisfaction and prosocial voice belief was also positive and significant (b = .14, SE = .07, p < .05). The ΔR2 for this step was significant (ΔR2 = .07, p < .001).

To further investigate the moderating effect of prosocial voice belief, we combined Aiken and West’s (1991) method and the Johnson–Neyman technique to examine the simple slopes of the regression curves at a range of job satisfaction values for the low (one standard deviation below the mean) and high (one standard deviation above the mean) levels of prosocial voice belief. As Table 5 shows, when prosocial voice belief was low, the simple slope of the curve was significantly negative when job satisfaction was 1.95 or less (b = −.35, SE = .18, p = .05). The simple slope was significantly positive when the value of job satisfaction was 5.21 or greater (b = .30, SE = .06, p = .05). For high prosocial voice belief, although the simple slopes were negative when the value of job satisfaction was 2.5 or less, they were statistically nonsignificant. The slopes were significantly positive when the value of job satisfaction was 3.55 or greater (b = .17, SE = .09, p = .05). As Fig. 2 illustrates, the results were consistent with our expectations; when prosocial voice belief was high rather than low, the negative relationship between job satisfaction and voice behavior was weaker at low levels of job satisfaction, and their positive relationship was stronger at high levels of job satisfaction. Thus, Hypothesis 2 was supported.

Table 5 Simple slopes of the relationship between job satisfaction and voice behavior for a range of job satisfaction values at high and low levels of prosocial voice belief
Full size table
Fig. 2
figure 2

The moderating effect of prosocial voice belief on the relationship between job satisfaction and voice behavior

Full size image

Hypothesis 3 involves the moderating effect of self-protective voice belief on the curvilinear relationship between job satisfaction and voice. As Model 5 in Table 3 shows, self-protective voice belief interacted with the squared term of job satisfaction (b = −.21, SE = .09, p < .05). To further probe the moderating effect, we examined the simple quadratic relationship between job satisfaction and voice behavior at one standard deviation above and below the mean of self-protective voice behavior (Aiken & West, 1991). When self-protective voice belief was strong, the coefficient of squared job satisfaction was not significant (b = .04, SE = .06, n.s.). By contrast, when self-protective voice belief was weak, we found a U-shaped curvilinear relationship between job satisfaction and voice behavior, as indicated by the positive and significant coefficient of squared job satisfaction (b = .36, SE = .09, p < .001). Figure 3 depicts this relationship, showing that the U-shaped curvilinear relationship between job satisfaction and voice occurred when self-protective voice belief was weak but not when it was strong. Thus, Hypothesis 3 was supported.

Fig. 3
figure 3

The moderating effect of self-protective voice belief on the relationship between job satisfaction and voice behavior

Full size image

Supplementary data analysis

We found a U-shaped relationship between job satisfaction and employee voice in this study. As the participants of this study were quite homogeneous in terms of age, education level, tenure, and geographic location, we collected additional data from a more diverse sample to cross-validate our finding about the curvilinear relationship. To obtain a diverse sample (Gosling, Vazire, Srivastava, & John, 2004), we recruited working adults online via Sojump, Inc., an organization that provides professional services of sampling and online data collection in China. Four hundred and twenty one employees completed our survey, in which they reported their job satisfaction, voice behavior, and demographic characteristics.

All of the participants had more than 6 months of working experience, and 79% had worked for more than 5 years. The respondents were in different age groups: 18–22 years old (2%), 23–27 years old (18%), 28–32 years old (30%), 33–37 years old (27%), 39–42 years old (10%), 43–47 years old (8%), and more than 48 years old (6%). Fifty eight percent of the participants were males. Eight percent of them had obtained a Master’s degree or above, 68% had completed their university studies, 17% had completed higher diploma programs, and 6% had only finished studies at senior secondary schools. They were from different cities in China and working in various types of organizations (27% in state-owned organizations, 17% in foreign enterprises, 15% in sino-foreign joint ventures, and 41% in private organizations).

We measured job satifaction and voice behavior with the same scales as those used in our main study. Consistent with the results of the main study, job satisfaction was positively related to voice (b = .48, SE = .03, p < .001), and squared job satisfaction term was also positively related with voice (b = .11, SE = .02, p < .001). These results demonstrated the same U-shaped relationship as that found in the main study, suggesting that the findings found in our main study are generalizable to other samples.

Discussion

We found that there was a U-shaped curvilinear relationship between job satisfaction and voice behavior and that this relationship was moderated by two individual beliefs about voice. Specifically, prosocial voice belief affected the slope of the relationship, making the positive relationship between job satisfaction and voice at high levels of job satisfaction stronger and the negative relationship between job satisfaction and voice at low levels of job satisfaction weaker. Self-protective voice belief attenuated the curvilinearity of the relationship between job satisfaction and voice, such that the relationship was less U shaped when self-protective voice belief was stronger.

Theoretical implications

The findings have several important theoretical implications. First, the overall positive effect of job satisfaction on voice (see meta-analyses of Ng & Feldman, 2012; Thomas, Whitman, & Viswesvaran, 2010) and the assumption of the linear relationship between these two variables have drawn voice scholars’ attention away from the findings that employees may speak up to improve their condition in response to their low job satisfaction (Farrell, 1983; Hirschman, 1970; Tangirala & Ramanujam, 2008; Zhou & George, 2001). Integrating this voice motivation and the motivation to reciprocate the organization’s kindness with voice and conceptualizing the salience of these motivations across the levels of job satisfaction, we found that the relationship between job satisfaction and voice is U shaped, with the right (positive) tail of the curve more pronounced than the left (negative) tail (see Fig. 1). This more complete and accurate picture better explains why and when employees exercise voice (cf. Tangirala & Ramanujam, 2008).

Second, the curvilinear relationship between job satisfaction and voice also sheds light on the dual benefits (organizational and personal) of voice behavior. Employees may engage in organizational citizenship behavior to benefit both their organizations and themselves (Bolino, 1999; Grant & Mayer, 2009; Rioux & Penner, 2001). Although the primary motive of voice is prosocial (Liang et al., 2012; Stamper & Van Dyne, 2001; Van Dyne & LePine, 1998), employees who speak up can potentially enjoy various positive outcomes, such as making their jobs easier (Morrison, 2014), attracting attention and resources (Ashford et al., 1998), creating positive social images (Fuller et al., 2007), and gaining influence and control (Tangirala & Ramanujam, 2008). Employees understand these benefits and may take them into consideration when deciding whether to speak up or remain silent. However, previous research on the antecedents of voice behavior has mainly focused on employees’ prosocial motives and, by and large, ignored the potential benefits of voice behavior to the voicers themselves (Morrison, 2014). It is important to integrate the self-serving effect of voice with the literature because understanding it helps uncover the exact pattern of employee voice determinants. As Klaas, Olson-Buchanan, and Ward (2012) note, when researchers recognize that employees speak up to benefit both their organizations and themselves, they can detect different effects of previously studied factors such as job attitudes on employee voice. Our study also answers calls for research to integrate rather than dichotomize self-interest-based and other-oriented motives of organizational citizenship behavior (De Dreu & Nauta, 2009; Grant & Mayer, 2009).

Third, our study responds to Detert and Edmonson (2011) call for more “follower-centric models” in voice research. Employees’ beliefs are vital to explaining voice behavior, in addition to contextual factors. Although beliefs are not always reflective of reality, they influence how people interpret cues in the environment and themselves and react (Detert & Edmonson, 2011; Epitropaki & Martin, 2004). Therefore, examining employee beliefs, especially those related to voice, can enhance the understanding of why employees engage in or avoid voice behavior. Prosocial voice belief pertains to the appropriate beneficiary of voice, while self-protective voice beliefs involve employees’ internal knowledge structures about the cost of voice behavior (Detert & Edmonson, 2011). Our study reveals that these voice beliefs affect employees’ decision about whether to speak up at different levels of job satisfaction. Specifically, the relationship between job satisfaction and voice is more positive (less negative) for employees with a stronger prosocial voice belief and less significant for employees with a stronger self-protective voice belief.

Finally, our finding on the moderating effect of prosocial voice belief confirms the notion that norms influence individual behavior (Cialdini & Trost, 1998). Previous research into this notion has primarily focused on the main effects of norms (Cialdini et al., 1990; Garst et al., 1997; Schwartz, 1973). An exception is the finding that the negative reaction to bad events is attenuated when the events are supported by norms (Duffy, Ganster, Shaw, Pagon, & Johnson, 2006; Leung, Lin, & Lu, 2014). Our finding reveals another type of moderating effect of norms. Specifically, people have multiple motives, and personal normative beliefs determine which motive drives their behavior—that is, a motive consistent with their normative belief is more likely to influence their behavior. This insight adds to the existing knowledge on the norm–behavior relationship.

Practical implications

The U-shaped relationship between job satisfaction and voice behavior informs managers that the most satisfied and unsatisfied employees are most likely to speak up in organizations; however, they speak up for different reasons. To elicit voice from the most satisfied employees, managers should emphasize the importance and benefits of their constructive ideas and suggestions to the organization. By contrast, it is easier to elicit constructive voice from unsatisfied employees by emphasizing the value of voice behavior for improving their work conditions and reducing their job dissatisfaction. Focusing on any of these benefits may not motivate employees who fall in the middle range of job satisfaction to speak up, because they neither feel obligated to reciprocate nor desperate to change the status quo. To elicit voice from these people, managers need to develop a voice climate that stimulates voice behavior regardless of employees’ work attitudes (Morrison et al., 2011).

Managers should also pay close attention to employees’ beliefs about voice behavior. Stressing the benefits of voice to the organization is more effective in stimulating voice behavior among employees who have a stronger prosocial voice belief. Thus, managers should tailor their voice management strategies to employees with different levels of prosocial voice beliefs. Conversely, self-protective voice beliefs demotivate employees to speak up despite the potential organizational or personal benefits of doing so. Therefore, managers should endeavor to establish an environment in which self-protective voice belief is unlikely to develop. For example, they may achieve this by being open to employees’ suggestions and recognizing and rewarding those who speak up about issues important to the organization (Detert & Edmonson, 2011).

Limitations and future research directions

This study has several limitations that could be addressed in future research. We argued and found that there is a U-shaped relationship between job satisfaction and voice because employees’ primary motive for voice behavior varies at different levels of job satisfaction. Although our results are consistent with this argument and the moderating effects of two voice beliefs provide evidence for this process, we did not directly examine employees’ voice motives. Future research could operationalize employees’ motivation to reciprocate and motivation to address problems so as to regain satisfaction and examine whether these two factors mediate the curvilinear relationship between job satisfaction and voice.

To improve the response rate in our survey, we measured all the variables in our model through employees’ self-report. Although this approach may have led to concerns about common method variance, the results of CFA showed that the latent constructs were distinct and the one-factor model fitted the data very poorly, suggesting that the common method bias was not serious in our study. In addition, our major findings involve significant quadratic and interaction effects. As common method variance does not inflate the chance of detecting quadratic (Peng, Wong, & Song, 2016; Podsakoff, MacKenzie, & Podsakoff, 2012) or interaction effects (Siemsen, Roth, & Oliveira, 2010), our results are unlikely to be the artifact of common method bias. Our two-wave design should also alleviate the concern of common method variance to some extent. Nevertheless, future researchers may try to replicate our study using data on employees’ voice behavior from different sources, such as supervisors.

The respondents in our main study were newly hired employees. As newcomers are an active and crucial source of organizational innovation (Choi & Levine, 2004; Hansen & Levine, 2009), it is important to examine their voice behavior. However, newcomers have certain characteristics that distinguish them from more experienced employees. For example, they tend to have limited resources such as time, energy, and supportive relationships (Cooper-Thomas & Burke, 2012), and identify less with their organizations (Mael & Ashforth, 1992). While our supplementary data analysis with a sample with diverse tenure has also found the U-shaped relationship between job satisfaction and voice behavior, future research may consider examining whether all of our findings in the main study can be replicated among employees with longer organizational tenure. In addition, since culture may influence people’s belief and their motives for proactive behaviors in the workplace (e.g., Li, He, Yam, & Long, 2015), researchers may also test whether our results are generalizable to other cultural contexts in Asia.

Conclusion

We found a U-shaped curvilinear relationship between job satisfaction and voice, moderated by employees’ prosocial voice belief and self-protective voice belief. These relationships exist because employees may speak up to reciprocate or speak up to address problems so as to regain job satisfaction, and the relative salience of these voice motivations varies depending on the level of job satisfaction. This finding extends the voice literature by providing a more accurate explanation of when and why employees speak up. With this knowledge, managers will be able to tailor their strategies to elicit employees’ voice based on their work attitudes and voice beliefs.