First, this study suggests that job insecurity would increase the extent to which an employee hides knowledge (12, 15, 29). Knowledge hiding is defined as the deliberate concealment of knowledge when another employee requests information. Knowledge hiding makes it difficult to maintain an organization's competitive advantage and achieve success in a dynamic and rapidly changing organizational environment, because it prevents employees from sharing and transferring crucial work-related information, knowledge, and expertise (12–14). Although existing works have paid little attention to the association between job insecurity and knowledge hiding (15, 30), we rely on the conservation of resources theory (29) to suggest that job insecurity may increase the extent to which an employee hides knowledge. According to the conservation of resources theory (29), when an individual member faces the threat of losing resources, he or she is likely to attempt to reduce his or her energies and resources in the context or environment around him or her. Therefore, when an employee gets a sense of job insecurity, he or she is likely to redirect his or her energies and resources away from his or her tasks at work (15, 29). As a result, the employee would not make sufficient effort to share his or her knowledge with his or her colleagues.

In this paper, we expect job insecurity to reduce employees' psychological safety (3, 31–33). Psychological safety refers to “an individual's perception of being able to show himself or herself without fear of adversely affecting his or her status, career, or self-image” ((33), p. 708). Although research on the relationship between job insecurity and employees' psychological safety is scarce (3, 32), it is obvious that job insecurity has a detrimental effect on employees' psychological safety (31, 33). When employees find an organization psychologically safe, they engage in risk-taking behaviors and have less fear that their opinions or ideas will be rejected, which makes them voice these ideas and opinions (31). Consequently, employees do not hesitate to seek their coworkers or supervisors for support and feedback, because they believe that the latter will not treat them unfavorably. On the contrary, when employees consider an organization psychologically unsafe, they will feel pressured and afraid and will find it difficult to freely express their opinions and thoughts (31, 33, 34). This can further decrease their psychological safety. Thus, organizations with high job insecurity will have employees feel that they are not being respected and protected by their employer, which is likely to prevent them from raising issues and opinions beneficial to the organization's success and development (21, 35–37).

In this study, we suggest that lower levels of psychological safety would increase the extent of an employee's knowledge hiding (31, 33, 38). An employee who feels a low level of psychological safety is not likely to cooperate with his or her colleagues, nor actively share his or her opinions and create knowledge (39–41). Thus, when his or her degree of psychological safety is low, the employee will be reluctant to share ideas, thoughts, and feelings with his or her colleagues and he or she will likely find it difficult to ask for help due to the fear or anxiety of being criticized (42). Additionally, a low degree of psychological safety may lead the employee to form negative perceptions and be skeptical of interpersonal relationships within the organization. As a result, the employee is likely to perceive his or her colleagues as rivals for survival in this environment (38, 43). Therefore, employees with low levels of psychological safety would abstain from sharing their knowledge in their organizations.

Integrating the dynamics discussed above (i.e., job insecurity, psychological safety, and knowledge hiding), we suggest that employees' psychological safety will mediate the relationship between job insecurity and knowledge hiding. Our mediation model can be supported by a context-attitude-behavior framework (44, 45). According to this perspective, an organization is characterized by a number of environmental or contextual factors, such as systems, practices, rules, and climates, which mold employees' attitudes and behaviors. In employees' minds, job insecurity is a critical context that influences their attitudes, such as psychological safety, and eventually affects their behaviors, such as knowledge hiding. Thus, we suggest that psychological safety mediates the relationship between job insecurity and knowledge hiding.

Moreover, and more important, we suggest that servant leadership would positively moderate the relationship between Job insecurity and psychological safety (23, 46, 47). In other words, our research sets boundary conditions by focusing on servant leadership and its role in the relationship between job insecurity and psychological safety. Our argument that job insecurity may lower employees' psychological safety is reasonable and acceptable. However, job insecurity may not always affect psychological safety in the same way because in real organizations, several contextual/contingent factors (e.g., personality, gender, age, leadership style, organizational climate, rule, and systems) moderate the relationship between the two variables (23, 46, 47).

This paper focuses on servant leadership, which is one of many leadership styles. This concept can be defined as “an (1) other-oriented approach to leadership (2) manifested through one-on-one prioritizing of follower individual needs and interests, (3) and outward reorienting of their concern for self toward concern for others within the organization and the larger community” [(48), p. 114]. In particular, we suggest that servant leadership mitigates job insecurity's detrimental effect on psychological safety. Servant leadership would provide followers with effective guidance about how to properly cope with any negative emotions, personal problems, and crises that stem from job insecurity (49–53). This leadership style encourages employees to feel a sense of respect, mutual trust, support, and self-worth in the organization, eventually reducing job insecurity's negative impact on psychological safety (49, 50, 53). For instance, when a leader's servant leadership is high, his or her behavior helps followers effectively resolve their anxiety and fear of unstable employment, even if they feel very insecure in their jobs. As a result, the employee is less likely to feel less safe psychologically (48, 50, 53).

In contrast, when a leader's servant leadership is low, employees find it difficult to cope with the negative emotions, personal issues, and crises concomitant to their unstable jobs. They are also likely to feel less respected and supported by their leader (49, 50, 53). Thus, leaders' low servant leadership may induce the followers who suffer in their unstable jobs to feel that they cannot effectively cope with said jobs and to become isolated from the organization. The negative influence of job insecurity on psychological safety would thus fail to become adequately resolved and may even be amplified (48, 50, 53).

In sum, this study aims to understand how job insecurity influences knowledge hiding through the mechanism of employees' psychological safety. It predicts that servant leadership plays a moderating role in the link between job insecurity and psychological safety. Specifically, when servant leadership is high, job insecurity's negative impact is more modest than when it is low. We use structural equation modeling (SEM) to empirically test our proposed moderated mediation model. Our study contributes to and extends our knowledge on job insecurity in the following way. First, we focus on a knowledge-related variable (i.e., employees' knowledge hiding) that is affected by job insecurity. Second, we establish the underlying mechanism that links employees' job insecurity and knowledge hiding. Third, we propose a way in which organizations can attenuate the negative impact of job insecurity on employees' psychological safety by suggesting that servant leadership moderates the relationship between the two. Finally, we utilized three-wave time-lagged data, which minimizes the possibility of common method bias compared to a cross-sectional study. Figure 1 visualizes our hypothesized model.

We collected time-lagged data from three waves of surveys, with each new wave taking place 4 or 5 weeks after its predecessor. We used an online survey administration firm, which maintains approximately 3,450,000 panel data representatives of the Korean population, to recruit adults who work full time. We aim to minimize the common method variance problems inherent to cross-sectional studies by using 3-wave time-lagged data. The survey participants registered through an authentication system and had to identify their occupation status by providing an e-mail address or a mobile phone number. Existing research has established that using such an online survey system is a reliable way of obtaining diverse samples (54).

The operation function of the online system allowed us to ensure that the surveys were distributed to the same participants, with each new wave taking place 4 or 5 weeks after its predecessor. The survey system gave the participants sufficient time to respond to each survey (e.g., 2 or 3 days) and the respondents were allowed to answer the survey whenever they wanted.

In an effort to ensure an efficient response, the survey firm used traps and timestamps for geographic IP violators to monitor the integrity of the data. These tools prevented participants from repeatedly accessing the survey and completing the questionnaire. In order to guarantee the spontaneity and confidentiality of our participants' responses, the experts at the survey firm contacted them directly to request their consent to take part in our survey. We promised compliance with common ethical standards (such as ensuring anonymity) to those who agreed to participate, and received their informed consent. Our respondents received a cash reward equivalent to US$8. Our study received institutional review board (IRB) approval from our respective universities in South Korea.

A total of 1, 512 participants completed our first (Time 1) survey, designed to measure job insecurity and servant leadership. Four weeks later, 421 participants completed our second (Time 2) questionnaire, which measured employees' psychological safety. Five weeks later, the same participants received our third (Time 3) survey, which measured knowledge hiding. Thus, after excluding the missing data, out of 512 participants, our final analysis sample consisted of 365 participants who answered all three questionnaires (a response rate of 71.29%). We used several recommendations proposed by earlier studies to calculate our sample size. First, we used G^*Power version 3.1.9.7 to determine the minimum sample size in order to assess whether our sample size was appropriate. According to our power analysis, which is based on previous research (55), 365 different samples had the adequate power (0.80) to identify a medium effect at an alpha level of p = 0.05. Furthermore, our research model followed the rule of ten (56), which states that one observable variable should include at least 10 cases. Considering that our research model contained 24 observable variables, thus our 365 cases would be sufficient. Table 1 provides a description of the characteristics of our respondents.

We measured different variables at each of the three survey points. At time 1, we measured the degree of job insecurity and servant leadership. At time 2, employees were asked to report their psychological safety. Finally, at time 3, we measured the employees' knowledge hiding behavior from their direct supervisors. All variables were measured on five-point Likert scales (1 = strongly disagree, 5 = strongly agree). We also calculated the variables' internal consistency via their Cronbach alpha values.

We measured employees' job insecurity using five items developed by Kraimer et al. (57). A sample item read: “My job is not a secure one.” The value of the Cronbach's alpha is 0.91.

Servant leadership was measured by Liden et al. (58) seven items. A sample item is: “My supervisor puts my best interests ahead of his/her own”. The value of Cronbach's alpha is 0.79.

We measured employees' psychological safety using seven items (31). A sample item read: “I am able to bring up problems and tough issues in this organization.” The value of the Cronbach alpha is 0.80.

We utilized five items of knowledge hiding behavior scale which consists of eleven items (59). Each employee's immediate supervisor evaluated the level of his or her knowledge hiding behavior. The reason why we shortened the full items is that the five items were validated by existing empirical research which were conducted in the context of South Korea (60). A sample item is “This employee pretended that he or she couldn't find the information that his or her colleagues wanted”, and “Thie employee gives colleagues a little bit of assistance, but didn't help them to the extent they wanted”. The value of the Cronbach's alpha is 0.95.

We measured several control variables in addition to our primary variables of interest. In line with existing research on knowledge hiding (59), we used participants' gender, tenure, and education level, all measured in our Time 2 survey.

We first conducted a frequency analysis to confirm the demographic characteristics of our participants. The relationships among the variables in our study were calculated through correlation analysis using SPSS 26 and Pearson. We followed Anderson and Gerbing (61) by adopting a two-step procedure (i.e., a measurement model and a structural model). We then conducted a series of confirmatory factor analyses (CFA) to evaluate the empirical distinctiveness of our main variables (i.e., job insecurity, servant leadership, psychological safety, and knowledge hiding). Afterwards, we used AMOS 23 to run our structural model and performed an analysis of moderated mediation using the maximum likelihood (ML) estimator.

We calculated the comparative fit index (CFI), the Tucker-Lewis index (TLI), and the root mean square error of approximation (RMSEA) to ensure the empirical distinctiveness of each of our main variables. Browne and Cudeck (62) suggest that it is ideal to have CFI and TLI values above 0.90 and an RMSEA value below 0.06. Next, we ran a bootstrapping analysis to confirm the significance of the indirect effect of psychological safety (63). Finally, we performed a bootstrapping analysis by estimating a 95% confidence interval (CI) to see if our hypothesis of mediation and indirect mediation was supported. When the confidence interval (CI) excludes zero (0), the indirect effect is significant at level 0.05 (63).

We find that some of the variables in our study are significantly correlated with one another (e.g., job insecurity, servant leadership, psychological safety, and knowledge hiding). Table 2 presents the descriptive statistics of and the correlations between our variables.

We first conduct a series of CFAs to ensure that each of our main variables is empirically distinctive (job insecurity, servant leadership, psychological safety, and knowledge hiding). Specifically, the chi-square difference test allows us to establish whether our proposed four-factor model (job insecurity, psychological safety, servant leadership, and knowledge hiding) is better than alternative models (e.g., a three-factor model, a two-factor model, and a one-factor model).

First, the hypothesized 4-factor model had a good and acceptable fit (χ² (df = 109) = 212.224; CFI = 0.974; TLI = 0.967; RMSEA = 0.051). Then, we conducted a series of chi-square difference tests by comparing the 4-factor model with a 3-factor model (χ² (df = 112) = 1239.387; CFI = 0.715; TLI = 0.655; RMSEA = 0.165), a 2-factor model (χ² (df = 114) = 1699.100; CFI = 0.600; TLI = 0.523; RMSEA = 0.194), and a 1-factor model (χ² (df = 115) = 1763.298; CFI = 0.584; TLI = 0.508; RMSEA = 0.197). The results of the chi-square difference tests showed that the 4-factor model was better than others. Thus, this result means that our four research variables have an appropriate degree of discriminant validity.

Our study uses a moderated mediation model that adds mediators and moderators to the relationship between job insecurity and knowledge hiding. First, the mediator (psychological safety) mediates the relationship between job insecurity and knowledge hiding. Second, the moderator (servant leadership) ameliorates the negative relationship between job insecurity and psychological safety.

Next, we multiply job insecurity and servant leadership in the moderation structure to create an interaction term. The variables are mean-centered to avoid multicollinearity. This technique reduces the multicollinearity and the correlation between our two variables (64).

We calculate tolerance values and variance inflation factors (VIF) to assess the effects of multicollinearity (64). We find that the tolerance values of job insecurity and servant leadership are 1.000 and 1.000, respectively. Their VIF values are 1.001 and 1.001, respectively. The results demonstrate/suggest that the two variables (job insecurity and servant leadership) are free of multicollinearity, as their tolerance values are >0.2 and their VIF values are < 10.

We conduct a chi-square difference test that compares our full mediation model to a partial mediation model to identify the best mediation model. Except for the direct path from job insecurity to knowledge hiding behavior, the full and the partial mediation models are identical. The results of the fit indices are reasonable both for the full mediation model [χ² = 295.377 (df = 137), CFI = 0.957, TLI = 0.947, and RMSEA = 0.056] and for the partial mediation model [χ² = 287.111 (df = 136), CFI = 0.959, TLI = 0.949, and RMSEA = 0.055). Yet, the chi-square difference test between the two (Δχ² [1] = 8.266, p < 0.01) indicates that the partial mediation model is significantly better than the full mediation model. The findings suggest that job insecurity affects knowledge hiding behavior both “indirectly” (through psychological safety) and “directly”.

We add our control variables into our research model since the variables can affect the dependent variable (i.e., knowledge hiding behavior). And we found that all control variables are statistically insignificant.

Then, the results showed that job insecurity and knowledge hiding are positively and significantly correlated (β = 0.17, p < 0.01). Thus, Hypothesis 1 is supported. Hypothesis 1 expected the “partial” mediation model (which was superior to full mediation) to contain the coefficient value of the relationship between job insecurity and knowledge hiding behavior. These results are consistent with the fact that the partial mediation model's fit indices are superior to those of the full mediation model. When we take the two together, we accept the Hypothesis 1. That is, rather than only indirectly, job insecurity is more likely to both “directly” and “indirectly” affect knowledge hiding behavior via psychological safety.

We also discovered a significant negative association between job insecurity and psychological safety (β = −0.27, p < 0.001), supporting Hypothesis 2. In addition, Figure 2 indicates that psychological safety is negatively and significantly correlated with knowledge hiding behavior (β = −0.21, p < 0.001), supporting Hypothesis 3 (Please see Table 3).

Hypothesis 4 predicted that psychological safety mediates the impact of job insecurity on knowledge hiding. To test Hypothesis 4, we perform a bootstrapping analysis (sample size = 10,000) (63). We use the resulting bias-corrected confidence interval (CI) to determine whether the mediation is significant. The 95% confidence interval (CI) should exclude the zero for us to be able to declare the mediation significant (63). We use these guidelines and a sample of 10,000 to confirm that psychological safety's indirect effect is significant, as the confidence interval does not contain zero (95% confidence interval [0.020, 0.105]. Thus, the mediating effect of psychological safety is statistically significant, and Hypothesis 4 receives some support. Table 4 illustrates the direct, the indirect, and the total effect of job insecurity on knowledge hiding.

Hypothesis 5 proposed that servant leadership positively moderates the negative relationship between job insecurity and psychological safety. To test Hypothesis 5, we mean-centered the two variables and generated an interaction term (i.e., job insecurity × servant leadership). The results demonstrated that the interaction term is significantly related to psychological safety (β = 0.14, p < 0.01). As expected, servant leadership plays a buffering role in the negative relationship between job insecurity and psychological safety. Specifically, the negative effect of job insecurity on psychological safety is buffered when servant leadership is high compared to when it is low. Taken together, these findings support Hypothesis 5 (Please see Figure 3).

From a theoretical perspective, this study makes the following contributions. First, we have examined the relationship between job insecurity and knowledge-related behaviors. The relationships between job insecurity and behavioral outcomes (e.g., innovative behavior, voice behavior, organizational citizenship behavior, safety behavior, and counterproductive work behavior) are well-documented (15, 16, 18–20). However, surprisingly, the research on knowledge-related outcomes is considerably limited (3, 15). Given that knowledge is the driving force behind organizational innovation and ultimately determines firms' competitive advantage and success (12–14), it is necessary to examine the relationship between job insecurity and knowledge-related behaviors. Therefore, this study enriches the literature on job insecurity by engaging in this line of research.

Second, even though existing research has highlighted the need to explore additional underlying mechanisms and boundary conditions in the relationship between job insecurity and knowledge-related behaviors (3, 15), few studies have done so. To better understand why and when this relationship occurs, it is important to examine its mediators and moderators. By integrating a context-attitude-behavior framework with social identity theory, this study has emphasized the roles of employees' psychological safety as a mediating mechanism and servant leadership as a boundary condition. In doing so, we extend existing research on the relationship between job insecurity and knowledge hiding by adding a substantive intermediating mechanism and boundary condition to interpret how job insecurity affects knowledge hiding and when the impact of job insecurity is minimized or strengthened.

Third, existing research on job insecurity has established/discovered that leadership plays a critical role in buffering the negative impact of job insecurity (46). However, most studies have focused on individual-level variables, such as self-esteem, internal locus of control, proactive personality, psychological capital, resilience, and emotional intelligence (6, 17, 22–25), and or macro-level contextual moderators, such as labor market insecurity, social safety networks, and macro-economic conditions (3, 15). Previous studies (27, 28, 46) indicate that leadership is a critical factor in encouraging subordinates' perceptions of and attitudes and behaviors toward the organization. Thus, servant leadership functions as a pivotal contingent factor in the relationship between job insecurity and knowledge hiding via employees' psychological safety. Our moderated-mediation model highlights the essential role of servant leadership when examining the influence of job insecurity on knowledge hiding.

The results of our study also have some practical implications. First, they show that job insecurity has important implications for knowledge hiding. Our SEM reveals that job insecurity leads to increased knowledge hiding. Organizational managers should remember that job insecurity hinders the flow of knowledge across organization members because knowledge is crucial to the achievement of firms' organizational success and competitive advantage. Thus, reducing employees' job insecurity might prove more effective in preventing knowledge hiding than material or financial incentives. Organizations could implement human resource management practices like mentoring programs, long-term contracting with employees, and fair performance evaluations (46) to encourage such reductions.

Second, we suggest that psychological safety mediates the relationship between job insecurity and knowledge hiding. Reduced psychological safety can increase job insecurity's influence on knowledge hiding behaviors. Thus, implementing specific measures to fortify employees' organizational identification should be a concern for managers. They could increase firm reputation through firm activities, systems, or lectures that inspire employees to identify themselves as members of their organizations and form positive organizational images in their minds. Therefore, managers should strive not only to decrease job insecurity but also to increase employees' psychological safety.

Third, we propose that servant leadership buffers the negative impact of job insecurity on psychological safety. In particular, in today's rapidly changing business environment, it is necessary to guide, facilitate, and inspire employees to help solve, and cope with, difficulties within their organizations, such as job insecurity. By encouraging leaders to engage in servant behaviors via training systems and courses (e.g., emphasizing the importance of subordinate guidance, discovering subordinates' potential and growth, providing subordinates with opportunities to maximize their abilities), leaders can develop a servant leadership style.

We believe that the current study may meaningfully contribute to the literature on job insecurity and knowledge hiding behavior, but it still has some limitations that need to be addressed. First, this research could not measure employees' job insecurity in an objective manner because it only uses self-reported survey data, which is subjective. However, objective indicators, such as the downsizing rate, may not directly influence employees' perceptions and attitudes because these objective characteristic (e.g., downsizing rate) tend to be interpreted through each individual's sense-making processes, which means that the objective measure would be unconsciously reflected in each employee's responses. Thus, this paper suggests that future research needs to not only use both subjective and objective measures, but also needs to compare the differential effects of these different measures. Second, this research could not properly consider a number of external factors that can substantially affect employees' job insecurity. Numerous objective factors affect an employee's perception on his or her subjective job insecurity, such as companies' downsizing rates, the quality or characteristics of the human resource management system in place, and the features of the social insecurity system at the national level (60). Therefore, we suggest that future research should more fully consider the issue by controlling for such objective variables.

Third, the fundamental values and spirit that servant leadership pursues may be universal in in the West and in the East (65, 66). However, a number of cultural differences do affect individuals' understandings of the role leadership plays, and these differences eventually affect employees' responses to different leadership styles. South Korea has been affected by the Confucian hierarchical systems for the past several centuries, so Korean employees may be more familiar with a culture of command and discipline than their Western counterparts (65). As a result, Korean employees are likely to feel that their leaders' servant behaviors are not natural and effective in a real organization. Therefore, the results of this study should be interpreted with care.