Knowledge Hiding and the Hider's Innovative Behavior

Innovative behavior is ‘an employee's intentional introduction or application of new ideas within a work, role, group, or organization’ (Janssen et al., Reference Janssen, van de Vliert and West2004). It embodies activities such as seeking new technologies, suggesting new approaches to objectives, applying new work methods, and investigating and securing resources to implement new ideas (Yuan & Woodman, Reference Yuan and Woodman2010). For example, frontline workers could increase their productivity by production skill improvement and procedure optimization. They may also discover hidden quality problems during production, thereby generating and applying creative ideas to product design (Janssen, Reference Janssen2000). Theorization of Janssen (Reference Janssen2000) and Scott and Bruce (Reference Scott and Bruce1994) is later developed to conceptualize innovation as a complex behavior consisting of two phases: development and implementation of novel ideas (Yuan & Woodman, Reference Yuan and Woodman2010). Notably, the first phase includes not only generating novel ideas by oneself but also adopting those from others (Arain et al., Reference Arain, Bhatti, Hameed and Fang2019).

These definitions suggest that innovative behavior begins with the development of new ideas (Yuan & Woodman, Reference Yuan and Woodman2010), featuring a novel combination of varied perspectives and approaches (Perry-Smith, Reference Perry-Smith2006). It is apparent that this process will require information and knowledge about a problem and prior knowledge regarding the task at hand (Gong, Kim, Lee, & Zhu, Reference Gong, Kim, Lee and Zhu2013). However, the ticklish question is that knowledge hiders may reject knowledge requests from coworkers in order to defend their knowledge territories from infringements (Connelly et al., Reference Connelly, Zweig, Webster and Trougakos2012), thus missing the opportunity to collect information and become aware of problems in the organization. Such knowledge is in fact necessary for generating new solutions because what is requested often reflects problems and difficulties their coworkers are confronted with (Rhee & Choi, Reference Rhee and Choi2017). This suggests that knowledge hiding will lessen a hider's ability to generate novel ideas (Bogilović et al., Reference Bogilović, Černe and Škerlavaj2017). Furthermore, as Černe et al. (Reference Černe, Nerstad, Dysvik and Škerlavaj2014) have demonstrated, when knowledge hiding is perceived, being a target of it instigates reciprocal distrust in the perpetrator. Such reactions not only undermine interpersonal relationships but also prompt the target to retaliate by concealing knowledge from the hider (Burmeister et al., Reference Burmeister, Fasbender and Gerpott2019; Xiao & Cooke, Reference Xiao and Cooke2019). This in turn prevents the hider from introducing new ideas of the seeker (Černe, Hernaus, Dysvik, & Škerlavaj, Reference Černe, Hernaus, Dysvik and Škerlavaj2017). Therefore, one can expect knowledge hiding to inhibit the development of new ideas, the first stage toward innovation.

Moreover, novel ideas introduce uncertainty (Baer, Reference Baer2012), such that whether employees will engage in the risky endeavor of implementing novel ideas (Yuan & Woodman, Reference Yuan and Woodman2010) largely depends on their ability or social relationships that can be leveraged to secure organizational resources from supporters (Baer, Reference Baer2012). However, this form of resource acquisition may be difficult for those whose social relationships have deteriorated due to knowledge hiding (Burmeister et al., Reference Burmeister, Fasbender and Gerpott2019; Xiao & Cooke, Reference Xiao and Cooke2019). That is, knowledge hiding may hinder the implementation of novel ideas, the second stage toward innovation. Given that knowledge hiding suppresses both essential stages of innovative behavior, we predict that:

Mediating Role of Silence Behavior

Silence behavior refers to ‘the conscious withholding of seemingly meaningful information, ideas, questions, concerns, and suggestions about potentially important work-related issues’ (Morrison & Milliken, Reference Morrison and Milliken2000). Brinsfield (Reference Brinsfield2013) argued that silence behavior was a typical means of self-protection for employees; by acting (and speaking) with discretion, employees keep a low profile in the organization, thus avoiding retaliation, criticism, and negative impact on career. Although both behaviorally appearing as the reluctance to speak up in the workplace, silence behavior and knowledge hiding are two distinct constructs (Xiao & Cooke, Reference Xiao and Cooke2019). First, knowledge hiding only occurs upon the initiation of a clear work-related knowledge request, whereas silence behavior does not rely on this prerequisite. Second, while knowledge hiding can be engaged using several strategies, such as evasive hiding and rationalized hiding, silence behavior only manifests itself as simply reducing the expression of work-related information. Third, knowledge hiding typically only involves know-how, whereas silence behavior represents a broader range of content from complaints, concerns to suggestions about potential organizational problems.

Silence behavior is entrenched in collectivist countries (Jain, Reference Jain2015). Collectivism is highly valued in China and maverick activities are less appreciated in this culture (Earley, Reference Earley1989). In similar situations where dissenters will be the minority of the group, people may refrain from speaking their minds (Duan et al., Reference Duan, Bao, Huang and Brinsfield2018). Moreover, in the context of Chinese culture where relationships and ‘face’ are stressed and traditional wisdom says that ‘Silence is golden’ and ‘The nail that sticks out gets hammered down’, people tend to take into account the feelings of peers and superiors for deciding whether to speak up in the workplace (Bedford, Reference Bedford2011). If speaking up is likely to harm the ‘face’ of someone, they may prefer to remain silent. Some scholars put that employees tend to remain silent (Morrison & Milliken, Reference Morrison and Milliken2000), considering unfavorable results that voice could have (e.g., being disliked by managers, causing jealousy among peers for exposing knowledge). Therefore, for employees under the influence of Chinese culture, silence might be an implicit yet prevalent code of behavior and a major self-protection strategy (Wang, Wu, Liu, Hao, & Wu, Reference Wang, Wu, Liu, Hao and Wu2019).

The theory of territorial behavior suggests that individuals who perceive potential threats to their knowledge territories are inclined to reinforce anticipatory defense strategies prior to any infringement attempt (Brown et al., Reference Brown, Lawrence and Robinson2005). Knowledge hiding reflects the hider's strong feeling of psychologically exclusive ownership toward knowledge assets (Das & Chakraborty, Reference Das and Chakraborty2018; Peng, Reference Peng2013) and vigilance against infringements (Singh, Reference Singh2019). It is possible that the hider regards knowledge requests as a signal of covetousness for their knowledge (Jahanzeb et al., Reference Jahanzeb, Fatima, Bouckenooghe and Bashir2019; Serenko & Bontis, Reference Serenko and Bontis2016). Therefore, knowledge hiding is expected to trigger the adoption of an anticipatory defense strategy before a knowledge request even occurs. By remaining silent, the hider cuts attention he/she receives in the organization to the minimum, thereby reducing the possibility of being requested for knowledge (Brinsfield, Reference Brinsfield2013). We propose that knowledge hiding is likely to prompt the hider to adopt silence behavior as an anticipatory strategy in Chinese organizations, which renders him/her in a hidden state without needing to expose knowledge or answer a number of questions.

Furthermore, Maqbool, Černe, and Bortoluzzi (Reference Maqbool, Černe and Bortoluzzi2019) confirmed that silence behavior inhibited interaction of organizational members and hence was harmful to innovation. Employee innovative behavior includes problem discovery and innovative problem-solving and relies on reciprocal exchanges of knowledge and information (Scott & Bruce, Reference Scott and Bruce1994). While holding back opinions on problems of the organization, silent employees lose the opportunity to uncover problems of the organization and communicate with coworkers about possible solutions (Morrison & Milliken, Reference Morrison and Milliken2000). This hampers the hiders’ expansion of knowledge domains and the development of novel ideas, which indeed are sustained by knowledge diversity and inclusiveness (Baer, Reference Baer2012; Men et al., Reference Men, Fong, Luo, Zhong and Huo2019). Ultimately, knowledge hiding curbs the follow-up innovative processes of support seeking and innovative problem-solving (Baer, Reference Baer2012; Yuan & Woodman, Reference Yuan and Woodman2010).

In line with our argument, we theorize silence behavior as a mediator between knowledge hiding and innovative behavior. According to the theory of territorial behavior, knowledge hiding leads the hider to adopt anticipatory defense strategies (in our case, silence behavior) for safeguarding his/her territory and deterring potential threats to competitiveness (Brown et al., Reference Brown, Lawrence and Robinson2005; Das & Chakraborty, Reference Das and Chakraborty2018). Although silence behavior builds a boundary seemingly impermeable for infringements on knowledge territory, it also separates the hider from the rest of the organization, reduces his/her interaction with other organizational members, and consequently impairs innovation (Brown et al., Reference Brown, Lawrence and Robinson2005). We, therefore, propose the following assumption:

Zhongyong Thinking as a Moderator

Zhongyong thinking derives from Confucianism and has existed in China for over two thousand years (Pan & Sun, Reference Pan and Sun2018). Being one of the fundamental ways of thinking for Chinese people, it has a profound impact on life norms, meanings, and coping behaviors of the Chinese (Ji, Lee, & Guo, Reference Ji, Lee and Guo2010). Yang (Reference Yang2010) established that Zhongyong thinking was metacognitive in nature, encompassing skills of planning, monitoring, and evaluating progress. From a decision-making style perspective, Wu and Lin (Reference Wu and Lin2005) defined Zhongyong thinking as a behavioral choice to reach a balance between self-interest and collective interests. This deliberate choice is made following three principles: weighing and balancing, integrating, and harmony seeking. Weighing and balancing refer to the way an individual recognizes the dialectical relationships between contradictory elements and acknowledges that these elements can achieve a dynamic equilibrium through mutual complementation and mutual reinforcement. Integrating stands for how one should consider the needs and values of others and endeavor to build consensus. Harmony seeking means taking harmony as the code of behavior and adequately considering potential consequences in order to avoid extreme reactions. As the ultimate goal of holistic thinking and integration of diverse perspectives, harmony is at the core of Zhongyong thinking (Yang et al., Reference Yang, Zhang, Zhao, Zhao, Wang, Chen, Ding and Zhang2016).

Although silence behavior is an effective precaution to guard knowledge territories against infringements, it can be highly dysfunctional for organizations (Nechanska, Hughes, & Dundon, Reference Nechanska, Hughes and Dundon2020). Employee silence prevents timely identification and correction of problems that may cause damage to organizations (Brinsfield, Reference Brinsfield2013; Morrison & Milliken, Reference Morrison and Milliken2000). Having greater cognitive flexibility and a higher level of holistic thinking, Zhongyong-oriented employees are able to ascertain whether their behavior is situationally appropriate by stepping back to see the whole picture (Li, Reference Li2018). Hence, it is easier for such employees to understand that taking silence as a precaution could harm the organization. At the same time, to reconcile personal and organizational interests, such employees tend to take diverse perspectives, thus adopting strategies that help build consensus and derive a solution harmful to neither side (Ning, Omar, Ye, Ting, & Ning, Reference Ning, Omar, Ye, Ting and Ning2021; Yang et al., Reference Yang, Zhang, Zhao, Zhao, Wang, Chen, Ding and Zhang2016). Therefore, to the extent that Zhongyong-oriented employees strive for a ‘middle way’ between personal interests (e.g., knowledge territory protection) and organizational interests (e.g., business growth brought by problem-solving), they are not likely to use silence as an anticipatory defense strategy in order to hide knowledge.

Moreover, Zhongyong thinking disapproves of spontaneous decisions. Rather, it endorses deliberation on available cues from various perspectives, a trade-off between these cues, and inclusiveness of different possibilities (Cheung, Yang, Chiu, King, Chan, & Chan, Reference Cheung, Yang, Chiu, King, Chan and Chan2003; Pan & Sun, Reference Pan and Sun2018). High Zhongyong thinkers are able to devise multiple solutions to a problem based on multiple sources of information (Yang et al., Reference Yang, Zhang, Zhao, Zhao, Wang, Chen, Ding and Zhang2016). Zhou, Zhang, Li, Sun, and Luo (Reference Zhou, Zhang, Li, Sun and Luo2021) put that Zhongyong thinking facilitates flexible and integrated cognitive processing, which in turn brings creative solutions and changes by fostering creative thinking. Aside from simple and direct defense strategies such as silence behavior, we expect that Zhongyong-oriented employees can also produce other possible defense strategies to protect their knowledge territories, such as playing dumb, evasive hiding, and rationalized hiding (Connelly et al., Reference Connelly, Zweig, Webster and Trougakos2012; Singh, Reference Singh2019). In addition, as these employees pursue harmony and exercise creative problem-solving, they will recognize the course of action conducive to holistic harmony (Lin, Hong, Xiao, & Lian, Reference Lin, Hong, Xiao and Lian2020). Qu, Wu, Tang, Si, and Xia (Reference Qu, Wu, Tang, Si and Xia2018) stated that while expressing their appeals of interest, Zhongyong-oriented employees do not intend to disrupt the harmony between individualism and the organization. In this sense, a desirable defense strategy – which silence behavior is not – should guard personal knowledge territory against infringements without compromising collective interests. One such example is rationalized hiding, that is, to politely justify withholding the requested knowledge (Burmeister et al., Reference Burmeister, Fasbender and Gerpott2019). Consequently, even though sustained by a self-serving motive, knowledge hiding is not likely to induce Zhongyong-oriented hiders to adopt the defense strategy of silence behavior. Instead, they may resort to rationalized hiding or other defense strategies that can balance the interests of all sides.

In contrast, low Zhongyong thinkers tend to see things from the lens of personal interests, in the face of conflicts between personal and organizational interests (Wu & Lin, Reference Wu and Lin2005). Knowledge hiding is a perpetrator's behavioral expression of psychologically exclusive ownership toward knowledge assets (Das & Chakraborty, Reference Das and Chakraborty2018; Peng, Reference Peng2013). As the theory of territorial behavior suggests, the stronger an individual's behavioral expression of psychological ownership, the more likely the individual to adopt direct and effective defense strategies to prevent infringements on his/her territory (Brown et al., Reference Brown, Lawrence and Robinson2005). Useful in blocking future knowledge requests, silence behavior is evidently a simple but effective anticipatory defense strategy for employees in Chinese organizations who choose personal benefits over organizational interests. Given that the hider with a low level of Zhongyong thinking is inclined to prioritize personal benefits, we infer that knowledge hiding, a strong behavioral expression of psychological ownership toward knowledge, will induce such hiders to engage in silence behavior. Thus, we predict that:

Together, Hypothesis 2 and Hypothesis 3 indicate the presence of a moderated mediation relationship (Preacher, Rucker, & Hayes, Reference Preacher, Rucker and Hayes2007) where Zhongyong thinking moderates the indirect effect of knowledge hiding on the hider's innovative behavior through silence behavior. In the first place, we have suggested that thanks to their holistic thinking, greater cognitive flexibility, and creative problem-solving, employees with a high level of Zhongyong thinking can recognize the harm of silence behavior, develop wide-ranging defense strategies to protect knowledge territories from infringements, and choose strategies suitable for maintaining harmony among all sides. Therefore, though sustained by a self-serving motive, knowledge hiding is not likely to lead Zhongyong-oriented hiders to use silence behavior as an anticipatory defense strategy, because silence can be detrimental to the organization. This lower tendency of silence behavior will strengthen the ability to develop novel ideas (Zare & Flinchbaugh, Reference Zare and Flinchbaugh2019), thereby facilitating innovative behavior. Conversely, for employees with a low level of Zhongyong thinking, personal interests generally seem to be more important than organizational interests. They are more likely to engage in silence behavior, a simple, effective anticipatory defense strategy, in order to hide knowledge. This stronger tendency of silence behavior will, in turn, impede innovative behavior (Madrid, Patterson, & Leiva, Reference Madrid, Patterson and Leiva2015). In other words, the level of Zhongyong thinking determines the magnitude of the indirect effect of knowledge hiding on innovative behavior through silence behavior. Therefore, the following moderated mediation relationship is hypothesized:

Discriminant Validity Analysis

Confirmatory factor analyses (CFAs) were performed using Mplus 7.0 software to ensure our focal constructs (knowledge hiding, silence behavior, Zhongyong thinking, and innovative behavior) had satisfactory discriminant validity. Item parceling was used to reduce observed indicators so as to improve the reliability of indicators and overcome non-convergence issues (Nasser & Wisenbaker, Reference Nasser and Wisenbaker2003). We compared the hypothesized 4-factor model with a series of nested 3-factor models and a 1-factor model. The CFAs show that our hypothesized 4-factor model fitted the data better than other alternative models (see Table 1), providing evidence of the construct distinctiveness. In particular, RMSEA (root mean square error of approximation) is an important indicator of goodness of fit used to compare the disparity between the hypothesized measurement model and the saturated model that provides a perfect fit. The greater the discrepancy, the less desirable the hypothesized measurement model (Hu & Bentler, Reference Hu and Bentler1999). The desirable goodness of fit of the hypothesized 4-factor measurement model is reflected by its RMSEA being smaller than that of the other three alternative models (see Table 1).

Table 1.

Confirmatory factor analysis results for model comparison

Notes: KH = knowledge hiding; SI = silence behavior; ZY = Zhongyong thinking; IB = innovative behavior. SRMR = standardized root mean square residual. CFI = comparative-fit index. N = 239. **p < 0.01.

To further examine the distinctiveness of knowledge hiding and silence behavior, we compared correlations between each of these variables with innovative behavior. Evidence for discriminant validity would be established if the two correlations were unequal (Cohen & Cohen, Reference Cohen and Cohen1983). Using the online utilizer (Lee & Preacher, Reference Lee and Preacher2013), the z-score was −1.99 for the difference between knowledge hiding–innovative behavior correlation and silence behavior–innovative behavior correlation (p < 0.05, two-tailed). This figure, together with the CFA results, supports the notion that knowledge hiding and silence behavior are distinguishable constructs.

Descriptive Statistics

Table 2 presents the means, standard deviations, and correlation coefficients of the variables. Knowledge hiding is positively related to silence behavior (r = 0.34, p < 0.01) and negatively related to innovative behavior (r = −0.36, p < 0.01). Silence behavior is negatively related to innovative behavior (r = −0.22, p < 0.01). According to Bosco, Aguinis, Singh, Field, and Pierce (Reference Bosco, Aguinis, Singh, Field and Pierce2015), large effect sizes involving behaviors (knowledge hiding – silence behavior and knowledge hiding – innovative behavior in our study) are greater than roughly |r| = 0.25, and a medium effect size involving behaviors (silence behavior – innovative behavior in our study) is between roughly |r| = 0.10 and 0.25.

Table 2.

Zero-order correlations and descriptive statistics

Notes: N = 239. Gender was coded as: 1 = male and 2 = female. Age was coded as: 1 = 20–25 years of age, 2 = 26–30 years of age, …, 8 = 56–60 years of age, and 9 = 61 years of age or greater. Education was coded as: 1 = junior school, 2 = high school, 3 = college's degree, 4 = Bachelor's degree, and 5 = Master's degree or above. Positional level was coded as: 1 = frontline employee, 2 = frontline manager, 3 = middle-level manager, and 4 = high-level manager. Organizational tenure unit in years. Reliability on the diagonal. *p < 0.05, **p < 0.01.

Hypothesis Testing

Hypotheses 1–3 were tested through multiple regression analysis. The moderated mediation (Hypothesis 4) was tested using PROCESS, an SPSS macro (Hayes, Reference Hayes2012; Preacher et al., Reference Preacher, Rucker and Hayes2007). We examined multicollinearity by the variance inflation factors (VIF). As shown in Table 3, VIF scores are below the cut-off of 10.0 proposed by Ryan (Reference Ryan1997) and the more conservative cut-off of 5 proposed by Ringle, Wende, and Becker (Reference Ringle, Wende and Becker2015), indicating that multicollinearity was not an issue in this study. Table 3 lists the ordinary least square regression results for the hypotheses.

Table 3.

Hierarchical regression analysis

Notes: VIF = variance inflation factor. Interaction = Knowledge hiding × Zhongyong thinking. N = 239. †<0.1, *p < 0.05, **p < .01.

Model 5 in Table 3 shows that gender (B = 0.11, SE = 0.12, p = 0.38), age (B = −0.06, SE = 0.05, p = 0.19), positional level (B = 0.12, SE = 0.07, p = 0.10), and organizational tenure (B = 0.02, SE = 0.01, p = 0.11) were not significantly associated with innovative behavior. However, education (B = 0.35, SE = 0.07, p < 0.01), intrinsic interest in innovation (B = 0.19, SE = 0.06, p < 0.01), and organizational climate (B = 0.26, SE = 0.08, p < 0.01) were significantly associated with innovative behavior. Collectively, these control variables accounted for 31% of variance in innovative behavior. Furthermore, consistent with Hypothesis 1, knowledge hiding had a significantly negative association with the hider's innovative behavior (B = −0.26, SE = 0.10, p < 0.01) and accounted for 2% of incremental variance in innovative behavior.

Hypothesis 2 predicts silence behavior will mediate the relationship between knowledge hiding and innovative behavior. This mediation was examined by following the suggestion of Baron and Kenny (Reference Baron and Kenny1986). First, Model 5 in Table 3 shows that knowledge hiding is negatively associated with innovative behavior (B = −0.26, SE = 0.10, p < 0.01; Model 5). Second, Model 2 in Table 3 shows that gender (B = 0.04, SE = 0.13, p = 0.79), age (B = −0.05, SE = 0.05, p = 0.36), education (B = 0.00, SE = 0.07, p = 0.98), positional level (B = 0.04, SE = 0.08, p = 0.58), and organizational climate (B = 0.06, SE = 0.09, p = 0.52) were not significantly associated with silence behavior. However, organizational tenure (B = 0.03, SE = 0.01, p < 0.05) and intrinsic interest in innovation (B = −0.14, SE = 0.06, p < 0.05) were significantly associated with silence behavior. These control variables accounted for 12% of variance in silence behavior. Furthermore, knowledge hiding was positively associated with silence behavior (B = 0.41, SE = 0.11, p < 0.01) and accounted for 6% of incremental variance in silence behavior. Lastly, after regressing innovative behavior on silence behavior, we found that silence behavior had a significant, negative relationship with innovative behavior (B = −0.10, SE = 0.06, p < 0.01; Model 6), but the effect of knowledge hiding on innovative behavior was reduced (B = −0.21, SE = 0.10, p < 0.05; Model 6). This means that silence behavior partially mediates the relationship between knowledge hiding and innovative behavior.

Moreover, a PROCESS analysis with 5,000 bootstrap samples was performed to further calculate estimates of the magnitude of direct and indirect effects. The magnitude of the direct effect of knowledge hiding on innovative behavior was −0.2148 (SE = 0.1024), at the 95% CI of [−0.4165, −0.0131]. The magnitude of the indirect effect of knowledge hiding on innovative behavior through silence behavior was −0.0445 (SE = 0.0261), at the 95% CI of [−0.1028, −0.0008]. These results indicate that Hypothesis 2 is supported.

Hypothesis 3 postulates that Zhongyong thinking will moderate the relationship between knowledge hiding and silence behavior. To test the moderating effect, control variables (gender, age, education, positional level, tenure, organizational climate, and intrinsic interest in innovation) were added into the regression in the first step. In the second step, knowledge hiding and Zhongyong thinking were entered, and the interaction term of these two variables was entered in the final step. Model 3 in Table 3 indicates the interaction between knowledge hiding and Zhongyong thinking is significant (B = −0.65, SE = 0.17, p < 0.01). To discern the direction of Hypothesis 3, a simple slope analysis was conducted on high (1 SD above the mean) and low (1 SD below the mean) levels of Zhongyong thinking (Aiken, West, & Reno, Reference Aiken, West and Reno1991). Figure 2 illustrates that knowledge hiding is positively associated with silence behavior at the low level of Zhongyong thinking (slope = 1.043, t = 5.391, p < 0.01), but not at the high level of Zhongyong thinking (slope = 0.097, t = 0.718, ns). Hypothesis 3 is supported.

Figure 2.

Interaction effect of knowledge hiding and Zhongyong thinking on silence behavior

Hypothesis 4 proposes that Zhongyong thinking will moderate the indirect relationship between knowledge hiding and innovative behavior through silence behavior. A PROCESS analysis with 5,000 bootstrap samples was performed to test the moderated mediation, and the indirect effect of knowledge hiding on innovative behavior through silence behavior was analyzed at two levels of Zhongyong thinking (1 SD above and below the mean). Table 4 indicates the conditional indirect relationship is significant for both 1 SD below the mean (indirect effect = −0.1022, SE = 0.0478, 95% CI [−0.2061, −0.0140]) and 1 SD above the mean (indirect effect = −0.0421, SE = 0.0253, 95% CI [−0.1004, −0.0024]). Overall, our results support Hypothesis 4.

Table 4.

Conditional indirect effect of knowledge hiding on innovative behavior

Notes: Boot LCI = bootstrapped lower confidence interval; Boot UCI = bootstrapped upper confidence interval.

Test of Endogeneity Bias

To find out whether the model is robust to potential endogeneity bias, we searched for instrumental variables (Antonakis, Bendahan, Jacquart, & Lalive, Reference Antonakis, Bendahan, Jacquart and Lalive2010) and specified two alternative models for mediation analysis, as recommended by Shaver (Reference Shaver2005). First, we allowed for the error correlation between knowledge hiding and silence behavior in the first alternative model. To keep the model identified, knowledge complexity served as the instrumental variable for knowledge hiding, because it was directly associated with knowledge hiding but not with silence behavior (see Shaver, Reference Shaver2005). Connelly et al. (Reference Connelly, Zweig, Webster and Trougakos2012) used knowledge complexity as a proxy for the time and effort required to transfer knowledge to a colleague and suggested this characteristic of knowledge was positively related to knowledge hiding. Silence behavior was defined as the intentional withholding of potentially meaningful questions, suggestions, or concerns (Brinsfield, Reference Brinsfield2013; Tangirala & Ramanujam, Reference Tangirala and Ramanujam2008). At least conceptually, knowledge complexity should not be related to silence behavior, which stems from fears and implicit theories about the risks of speaking up (Morrison, See, & Pan, Reference Morrison, See and Pan2015). The results of the endogeneity bias test indicated that the error covariance was non-significant and the relationship between knowledge hiding and silence behavior remained positive and significant (B = 0.16, p < 0.05).

Analogous to the above procedures, the second alternative model allowed for the error correlation between silence behavior and innovative behavior. We used moral disengagement as the instrumental variable for silence behavior. While a positive association exists between moral disengagement and silence behavior (He, Peng, Zhao, & Estay, Reference He, Peng, Zhao and Estay2019), innovative behavior is an extra-role behavior that employees have no moral obligation to perform. Thus, at least conceptually, moral disengage should not be related to innovative behavior. In this model, error covariance was non-significant, and the relationship between silence behavior and innovative behavior remained negative and significant (B = −0.42, p < 0.01). Taken together, the results demonstrate a certain degree of robustness of our findings to endogeneity bias.