Article body

In an organizational context, and as highlighted in the literature (Bartol & Srivastava, 2002; Connelly, Zweig Webster & Trougakos, 2012), the term “knowledge” does encompass “the information, ideas, and expertise relevant for tasks performed by organizational members” (Connelly et al., page 65). Knowledge is one of the main resources of organizations (Bock et al, 2005) and is critical to their long-term performance (Nonaka and Takeuchi, 1995). Thus, knowledge has to be managed, knowledge sharing closely monitored and knowledge hiding behaviors discouraged.

Even though it cannot be imposed, knowledge sharing should be nurtured and facilitated in organizational settings (Anand & Walsh, 2016). For instance, knowledge sharing may sometimes facilitate collaboration and business in cross-cultural settings (Barrett & Oborn, 2010). In multinational firms, where business is conducted on the international scene, culturally diverse colleagues have to work together, and managers have to guide international teams that are culturally heterogeneous. This cultural diversity can stimulate creativity at work, but only if organizational members do not hide their knowledge from each other (Bogilović, Černe, & Škerlavaj, 2017). Yet, despite the many efforts to foster knowledge sharing within organizations, employees often hide their knowledge. Even though not actively and purposely hiding knowledge does not always imply sharing it, knowledge can definitely not be shared if it is hidden. Thus, knowledge hiding should be avoided in organizations and knowledge hiding and its possible causes have been receiving increasing attention from researchers and practitioners alike (Burmeister et al., 2018; Connelly et al., 2012). However, variables that may actually discourage knowledge hiding in organizations have attracted less attention, and this is the focus of the present article. In order to help understand knowledge management in organizations, and how knowledge hiding may be discouraged, we investigate in this paper some relationships between employees, between employees and supervisor, and between the employees and the organization, through a survey administered in knowledge-intensive firms.

Supervisors play a role in the willingness of employees to share their knowledge (Kim et al., 2015; Kim et al., 2016) or hide their knowledge (Ladan et al., 2017; Lanke, 2018). The literature has investigated for various purposes, different facets of supervision as perceived by organizational members, e.g., abusive supervision (Connelly et al., 2012), supportive supervision (Anderson et al., 2002), and fair supervision (Colquitt, 2001). It has examined the negative effects of abusive supervision on employees’ knowledge sharing (Lee et al., 2017; Kim et al., 2015; Kim et al., 2016; Kim and Yun, 2015; Wu and Lee, 2016) and its influence on employees’ knowledge hiding (Khalid et al., 2018; Ladan et al., 2017; Lanke, 2018). However, non-abusive, supportive and fair supervision have never been studied together as the dimensions of a single construct and little is known about whether these three facets have actual adverse effect on employees’ knowledge-hiding behaviors. In this article, we model what we name “positively perceived supervision” as a multi-dimensional first-degree reflective, second-degree formative construct (Edwards, 2001; Law, Wong & Mobley, 1998) and show the significance of this construct for discouraging knowledge hiding within organizations. As co-worker and organizational support have also been shown to reduce the tendency of employees not to share knowledge while abused by their supervisors (Kim et al., 2015; Kim et al., 2016; Lee et al., 2017), we also test the possible adverse influence of these constructs on knowledge hiding.

This article sets out to answers the following research questions:

  1. Does positively perceived supervision discourage employee knowledge hiding behaviors?

  2. Does organizational support and co-worker support reduce the propensity of employees to hide knowledge?

The article is organized as follows. Using the literature, we first define the various constructs investigated and their possible relationships, leading us to lay down our hypotheses. We then describe our methodology and provide our results, before discussing them, their managerial implications, their limitations, and concluding.

The Various Constructs of Our Research

In this section, we investigate the various concepts/constructs called upon in our research and we lay down our hypotheses.

Knowledge Hiding

Organizations make efforts to facilitate knowledge-sharing activities to increase the performance of individuals and gain competitive advantage (Wang and Noe, 2010). However, not all individuals are willing to share their knowledge in every situation (Anand and Walsh, 2016). Some researchers suggest that people can hide knowledge from others due to personality characteristics, intrinsic values, or situational constraints (Anand et al., 2019; Connelly et al., 2012). Research also suggests that the propensity not to share and to hide knowledge may result mainly from interpersonal dynamics (Connelly et al., 2012). In this context, researchers have defined “knowledge hiding” as a theoretical construct to describe people’s behavior in resisting the sharing of knowledge with others. Knowledge hiding is “an intentional attempt by an individual to withhold or conceal task information, ideas, and know-how that has been requested by another person” (Connelly et al., 2012, p. 65). In organizations, knowledge hiding can significantly damage relationships at work, create distrust among co-workers, result in knowledge gaps, and lead to lower individual and organizational performance (Hernaus et al., 2019).

Connelly et al.’s (2012) significant work that defined and modeled knowledge hiding, found that it is not just the refusal to share knowledge but is a multi-dimensional construct comprising three factors that explain the behaviors behind hiding: rationalized hiding, evasive hiding and playing dumb (See Figure 1).

Figure 1

Knowledge Hiding

Knowledge Hiding
Based on Connelly et al., 2012

-> See the list of figures

In rationalized hiding, hiders provide a justification for failing to provide the knowledge requested by others. They justify this behavior with a convincing reason or put the blame on another party. In evasive hiding, hiders try to protect their knowledge by providing either incorrect information or a misleading promise to fulfill the request in the future, even though there is no intention to do so (Connelly et al., 2012). Playing dumb includes behaviors through which hiders pretend to be ignorant, or simply behave as if they are unfamiliar with the knowledge that is being required (Connelly et al., 2012). Using Connelly el al. (2012) measures of knowledge hiding, we investigate in the present study how this phenomenon may be discouraged However, we argue that within organizations employees might be differently inclined to hide their knowledge from different categories of people. For instance, organizational members, who perceive their supervisor as being abusive may be inclined to withhold knowledge from that supervisor but willing to share it with co-workers, particularly if they perceive them are supportive. Therefore, we argue that hiding knowledge from supervisors and co-workers should be considered and assessed separately (See Figure 2).

Positively Perceived Supervision

The literature highlights different aspects of supervisors’ behavior and impact on employees and suggests that supervisors can play a role in the willingness of employees to share their knowledge (Kim et al., 2015; Kim et al., 2016) or hide their knowledge (Ladan et al., 2017; Lanke, 2018). We argue that knowledge hiding is a factorially complex outcome that requires “predictors that are also factorially complex” (Edwards, 2001: 149). We study in this section, several aspects of supervision that have been investigated separately in the literature to explain other phenomena and could have an impact on employee’s knowledge hiding behaviors.

Employees often perceive their supervisors as rather abusive or non-abusive, based on their behavior toward them. Abusive supervisors are perceived as displaying characteristics such as ridiculing, humiliating, discouraging, and creating hostility (Tepper, 2007). Supervisors’ abusive behavior can cause employees to reduce their organizational commitment, decrease their job performance and increase counterproductive behavior (Connelly et al., 2012; Kim et al., 2015; Zhang et al., 2019). When employees are abused by their supervisors, they are more likely to quit the organization, have increased job dissatisfaction, and develop psychological distress (Tepper, 2000). Wu and Lee (2016) found that abusive supervision is negatively related to employee knowledge-sharing behavior and that employees are more likely to share knowledge with co-workers and not with supervisors while abused. In their study, Kim et al. (2016) found that abusive supervision discourages employees from sharing knowledge with others and has an effect on employees’ job-related performance. When employees are abused by their supervisor, they may resist knowledge sharing (Kim et al., 2015; Kim et al., 2016; Lee et al., 2017), and abusive supervision leads to active employee knowledge-hiding behaviors (Khalid et al., 2018). Employees perceive their knowledge base to be valuable and the feeling of being “mistreated or not given due respect” by supervisors will encourage them toward knowledge-hiding behaviors (Kim et al., 2016). Finally, when employees perceive that they are not being treated well by their supervisor, they may seek revenge by withholding knowledge from safe and easy targets (co-workers) (Khalid et al., 2018; Mitchell and Ambrose, 2007; Skarlicki and Folger, 1997). Supervision research in the literature has focused mostly on destructive consequences of abuse (Lee et al., 2017).

Figure 2

Knowledge hiding from supervisor and from co-workers to be assessed separately

Knowledge hiding from supervisor and from co-workers to be assessed separately

-> See the list of figures

Supervisors are sometimes unfair in their attitudes to the contributions made by employees in organizations (Holten et al., 2016). Employee satisfaction in organizations depends on the rewards and resources that are fairly distributed (distributive justice) in response to their contribution to organizational success (Colquitt, 2001; Ibragimova et al., 2012; Leventhal, 1976). “Distributive justice” is the term used to describe employees’ perceptions of fairness in recognition of their contribution to organizational expectations (Colquitt, 2001; Cropanzano and Ambrose, 2001; Deutsch, 1975). According to Colquitt (2001) and Holten et al. (2016), distributive justice can be witnessed through the outcomes that employees receive from their work (e.g., pay raises, promotions). Distributive justice is fostered when organizations allocate equity or equality (rewards and resources distributed in accordance with recipients’ contributions) norms for employees, and allocations can maximize productivity and improve cooperation (Leventhal, 1976). Thus, distributive justice (Colquitt, 2001), as distributed through supervisors, may be understood as fair supervision and has been shown to lead to employee satisfaction.

Perceived supervisor support is defined as the general views that are developed by employees concerning the degree to which supervisors value their contributions and care about their well-being (Eisenberger et al., 2002). If a supervisor’s behavior is perceived as supportive, it may help increase employee performance, provide strong emotional support, improve job satisfaction, and develop strong work relations (Anderson et al., 2002; Gonzalez-Morales et al., 2016; Yagil, 2006). Employees who feel supported by their supervisor perform better, are more committed to their employer (Frear et al., 2018), and are likely to demonstrate high levels of job-related performance (Kim et al., 2015). Some researchers have proposed that supervisor support helps to reduce stress, increase positive attitudes and behaviors (Carlson and Perrewe, 1999), and increase affective commitment (Casper et al., 2011). predicts several positive employee and work outcomes (Dysvik et al., 2014). Supportive supervision includes a type of social support which reduces the negative effects of stress and strain for employees, and supportive leaders are those who provide emotional, informational, instrumental, and appraisal support to followers (House, 1981; Paustian-Underdahl et al., 2013). Moreover, those employees who receive high levels of supervisor support tend to increase their efforts to exceed their responsibilities in return for the benefits provided by their supervisors (Shanock and Eisenberger, 2006). Supportive supervision is also understood as supportive leadership (Paustian-Underdahl et al., 2013). Ladan et al. (2017, p. 60) proposed that, when employees hide knowledge, transformational leadership through psychological ownership of knowledge may influence employees to refrain from knowledge-hiding behaviors that may negatively affect the organization and encourage knowledge sharing to improve the performance of the organization. Anderson et al. (2002) envisages and assesses supportive supervision as work–family balance facilitated by supervisors through supportive policies.

We argue that although supervisors may display non-abusive behaviors toward an employee, they may also, somewhat independently, exhibit fairness or unfairness in appreciating and recognizing the contributions made by that employee as well as demonstrating supportive behaviors related to work-family balance. Thus, two supervisors with different assessments of fairness, non-abusive behaviors and support may be equally positively perceived. Therefore, and beyond each single of the three highlighted dimensions, we argue that it is the global perception of supervision that appears important to assess, in terms of impact on knowledge hiding behaviors: each dimension could be extremely significant when considered individually on its own but less so when investigated as part of a whole as it could be counteracted by another dimension. Thus, we propose to study the impact of supervision on employees’ knowledge hiding behaviors through a construct that we name “positively perceived supervision”, which includes three complementary facets / dimensions: non-abusive supervision, fair supervision and supportive supervision (see Figure 3). Contrary to other studies, which investigate some dimensions of perceived supervision as unidimensional constructs, referring to single separate theoretical concepts (Hattie, 1985), we investigate the concept of positively perceived supervision as a single overall aggregate construct, which is a “composite of its dimensions, meaning the dimensions combine to produce the construct” (Edwards, 1998: 147). Variation in a multidimensional construct implies variation in any or all of its dimensions (Ibid.). Thus, the same assessment of such a construct may correspond to counteracting actions from its various dimensions. For instance, a supervisor who is non-abusive but also non-supportive of work family balance may score the same (with similar impact on employees’ knowledge hiding behaviors) as one who is abusive but also supportive of work family balance. Therefore, it appears essential to assess positively perceived supervision as an expression of the three identified relevant dimensions, and not only of one of these dimensions.

Hypothesis 1a: The more supervision is perceived positively by employees, the less they will be inclined to hide their knowledge from their supervisors.

Hypothesis 1b: The more supervision is perceived positively by employees, the less they will be inclined to hide their knowledge from their co-workers.

Figure 3

Positively perceived supervision

Positively perceived supervision

-> See the list of figures

Perceived Co-Worker Support and Organizational Support

As well as the supervisor, the organization and co-workers are major partners for employees and sources of social support in the workplace (Kim et al., 2015; Kim et al., 2016, Lee et al., 2017; Shoss et al., 2013). Ducharme and Martin (2000) suggest that the term “social support” helps to describe the relationship between work, co-workers, and well-being in a workplace. Co-worker support is a form of social support that comes from co-workers (Woo and Chelladurai, 2012). It is the most relevant form of social support for employees in the workplace (Kossek et al., 2011). It refers to employees’ beliefs about the extent to which co-workers provide desirable resources in the form of emotional support (e.g., showing concern) and instrumental support (e.g., helping with work tasks) (Xu et al., 2017). This support can be affective support and provide the recipient with feelings of being accepted and cared for; it can also be instrumental support and involve material assistance in response to specific needs (See Figure 4).

The availability of co-worker support may result in reduced turnover intentions and effective service recovery performance (Poon, 2011). Individuals who receive negative treatment from their supervisors may get support from their co-workers, their family, and their organization (Shoss et al., 2013). If supervisors are abusive, co-worker support becomes a more salient and important source of social support. Strong support from co-workers and supervisors improves the work environment by relieving employee stress (Poon, 2011). Co-worker support not only alleviates the negative effects of job characteristics that employees experience, it also enhances psychological well-being and the performance of employees (Sloan, 2012).

Hypothesis 2:The more employeesperceive support from co-workers, the less they will be inclined to hide knowledge from them.

Perceived organizational support refers to the general beliefs of employees about how much the organization values their contribution and cares about their well-being (Eisenberger et.al., 2002; Eser and Ensari, 2008) (See Figure 5).

Figure 4

Perceived Co-worker support

Perceived Co-worker support

-> See the list of figures

Figure 5

Perceived Organizational support

Perceived Organizational support

-> See the list of figures

Eisenberger et al., (1986) proposes that supervisor’s attitudes towards subordinates are indicators of organizational support because a supervisor, as an agent of the organization, has discretion and responsibility for managing and accessing subordinates’ performance. However, and although many studies do highlight the relationship between supervisory treatment of employees and perceived organizational support (Rhoades & Eisenberger, 2002), it has been also highlighted in the literature that an organization is mostly perceived as embodied by top management (Russo, Miraglia and Borgogni, 2013) whereas employees may vary in the extent to which they perceive their direct supervisors as embodying the organization (Shoss et al., 2013). Thus, perceived supervisor support and perceived organizational support may be considered as two separate construct and some scholar did differentiate them e.g., Rhoades & Eisenberger, 2002, Eisenberger et al., 2002; Hutchison, 1997; Kottke & Sharafinski, 1988). Hence, it can be argued that although employees may attach the relationships they have with their supervisors to their overall perceptions of the organization - only to the extent that supervisors are identified with the organization (Eisenberg, et al., 2002; Rhoades & Eisenberg, 2002) - employees can also perceive the two relationships as being separate and different: they may differentiate their perception of their supervisor from their experience with the organization as a whole. For instance, an employee might perceive her supervisor as supportive of her work-family balance through the authorization of fluctuating timetables that reflect children’s school schedules. The same employee might consider the organization as non- supportive of her work-family balance because of a general rule implemented within the organization that forbids more than three days off work to mind a sick child. Thus, without contravening organizational rules, the supervisor might be perceived by the employee as granting more flexibility. In this paper, we consider the construct of perceived supervision at the functional and operational level and based on person-to-person relationship, whereas organizational support is considered in a holistic perspective as a broad perception of the organization, an entity embedded in a set of rules and norms.

When employees feel supported by their organization, they develop a strong sense of commitment toward it (Woo and Chelladurai, 2012). Organizational support can reduce absenteeism and increase work performance. Concern and consideration by organizations can trigger various psychological mechanisms in employees. First, support from the organization should fulfill the socio-emotional needs of employees, enhancing their organizational membership. Second, perceived organizational support should reinforce employees’ perceptions that the organization appreciates and rewards good performance (Eser and Ensari, 2008; Rhoades and Eisenberger, 2002). Organizational support potentially increases the positive attitudes of employees at work, which may result in positive emotional associations with the organization itself, thereby increasing affective commitment (Eisenberger et al., 2010; Maertz et al., 2007) and decreasing turnover intentions (Mathieu and Zajac, 1990). Perceived organizational support affects employees’ perceptions that the organization cares about their well-being and values their contributions. Therefore, this is a situational factor that can influence employees’ attitudes to knowledge hiding and other work outcomes (Alnaimi and Rjoub, 2019). Furthermore, research has shown that organizational support in the workplace is positively related to knowledge sharing (Han et al., 2018; Yang et al., 2018)

Hypothesis 3a:The more employeesperceive support from their organization, the less they will be inclined to hide knowledge from their supervisor.

Hypothesis 3b:The more employeesperceive support from their organization, the less they will be inclined to hide knowledge from their co-workers.

All hypotheses are summarized in Figure 6.

Methodology

In this section we present our data collection procedures, our statistical approach and the measures we used for each construct investigated in our nomological network.

Data Collection and Treatment

The data were collected between December 2018 and March 2019 through an online survey. Even though anonymity was guaranteed in writing, the subject of perceived supervision is a sensitive topic, and we found that respondents were not always prepared to answer our questions and disclose the perception they had of their supervisors. Therefore, we used snowball sampling whereby we first sent the survey to employees, whom one of the authors knew in various organizations and who we felt could answer our survey without putting themselves in danger. We then asked each of these respondents to send the survey to people they knew and whom they could reassure about its purpose. This sampling method is deemed acceptable in hard-to-reach populations (Goodman, 2011).

As the explained variables of our model are related to knowledge hiding, we purposely sampled as first “seed” respondents, people who belong to knowledge intensive firms (consulting firms and higher education institutions) where knowledge management is essential and knowledge sharing usually broadly nurtured. We expected that many respondents would belong to the same institutions as the original respondents, which proved correct (see Appendix A). The “seed” respondents belong to three well-established higher education institutions (One French, one Indian and the third American, all middle-sized, employing between 200 and 500 employees, including administrative staff and professors) and one well-established large international French-based consulting firm (currently employing 4257 employees).

Figure 6

The research model

The research model

-> See the list of figures

We kept collecting data until the rules of thumb (Barclay et al., 1995; Cohen, 1992; Hair et al., 2017) were achieved. Both Barclay et al. (1995) and Hair et al. (2017) recommend that the sample size should be “ten times the largest number of structural paths directed at a particular construct in the structural path model”. Cohen (1992) recommends a sample size that takes into account a statistical power of 80%.

As proposed by Ringle, Sarstedt and Straub (2012), we used a partial-least-squares (PLS) approach and bootstrapping as a resampling technique (500 random samples) to generate t statistics (Chin, 1998), because our model was fairly complex (with several aggregate constructs), we did not have data with a normal distribution and had reflective and formative constructs in our model: PLS does not require data with a normal distribution (Fornell and Cha, 1994); it supports both reflective and formative constructs (Gefen et al., 2000).

We received 174 completed questionnaires. Five were discarded due to missing data in research-critical items, resulting in a final sample size of 169. This number of respondents and non-normal data are sufficient to perform partial-least-squares structural equation modeling (PLS-SEM) analysis. We used SmartPLS (Ringle, Wende and Becker, 2015) Version 3.2.6 for this study. The bootstrapping setting was tuned to 500 samples, using a bias-corrected and accelerated bootstrap with no sign changes and a two-tailed method. Demographic categorical variables (age, gender, industry, country) were included in the questionnaire, which allowed us to keep track of the population (see Appendix A for the characteristics related to our sample).

Measures

All measures were assessed using 5-point Likert scales with anchors of 1= Completely false or Never, depending on the question, and 5= Completely true or All the time, depending on the question. Details of all items are provided in Appendix B. We adapted Connelly et al.’s (2012) knowledge hiding scales to our context as they provide a comprehensive, three-dimensional (playing dumb, evasive hiding and rationalized hiding) measure that takes account of the different facets of this construct. Each reflective dimension was assessed through three items (see Appendix B), and we assessed separately employees’ knowledge hiding from their manager (KH_MAN) and from their co-workers (KH_CW).

For the positively perceived supervision (PS) measurement model, we drew from the literature and adapted to our context Peng et al.’s (2014) abusive supervision scale (ABUS_SUP: we reverse-coded respondents’ answers to obtain a score for non-abusive supervision[1]), Anderson et al.’s (2002) scale for supervision supportive of work–family balance (SUPP_SUP), and Colquitt’s (2001) and Leventhal’s (1976) scales for distributive justice (FAI_SUP: fair supervision). Each of these three dimensions was assessed with four items.

Finally, for perceived organizational support and perceived co-worker support, we employed the scales used by Woo and Chelladurai (2012). However, to adapt the scales to our context, we went back to the original scales used by Eisenberger et al (1986, 1997) for organizational support and Ducharme and Martin (2000) for co-worker support. Perceived organizational support was modeled with two dimensions (OS_VC: valuation of contribution and OS_WB: well-being), each assessed with five items. Perceived co-worker support was modeled with two dimensions (COWS_AFF: affective and COWS_IS: instrumental), each assessed with four items. Full details of the items and scales used are provided in Appendix B.

Results

Measurement Models

For all explanatory variables, which were aggregate second-order constructs (first-order reflective, second-order formative), i.e., PS, OS, and COWS, the indicator-re-use approach (Lohmöller, 1989; Ringle et al., 2012) was applied to test the measurement models; in this approach, all indicators of the lower-order components are re-used in the higher-order component. For the reflective dimensions of all measurement models, item loadings, reliability, and discriminant validity (if the constructs share more variance with their own measures than with other constructs, and if indicators load more strongly on their own constructs than on others) were confirmed through factor analysis.

Table 1 shows the cross-loadings for all reflective constructs. These should be greater than.707 for all items representing the same latent variable. If this is so, it shows that more than half of the variance is captured by the constructs. The loadings for the items of each block were similar in their representation of the underlying construct (see Table 1), confirming convergent validity. Discriminant validity was ensured through each construct square root of the Average Variance Extracted (AVE), which must be greater than its correlation with other factors (Gefen et al., 2000) (see Table 1). Reliability was assessed using Cronbach’s alphas and composite reliability (CR). Alphas should be greater than.70 (Nunnally, 1978) and CR should meet the minimum requirement of 0.7. AVE (convergent validity) greater than.50, meaning that 50% or more variance should be accounted for (Fornell, Claes and Larcker, 1981). All the criteria were largely met (see Table 2).

There were no collinearity issues among the predictor constructs: all VIF values were below 5 (see Appendix C) and all paths of the measurement models were significant (see Table 3).

For the explained variable knowledge hiding, which was also an aggregate second-order construct (first-order reflective, second-order formative), we used a two-stage approach in order to obtain the “true” AVE (Ringle et al., 2012). For this explained aggregate construct, the indicator re-use approach (Lohmöller, 1989) was applied during the first stage. Then, the latent variable scores for the lower-order components, obtained during the first stage, served as manifest variables for the higher-order components during the second stage. During this latter stage, the knowledge hiding construct was incorporated into the structural model.

Our results support the internal validity and reliability of the various scales that were used.

Structural Model

All the paths of our model were significant (see P values in Table 4). Hypotheses 1a, 1b and 2 are supported. However, surprisingly, the opposite of our hypotheses 3a and 3b is supported.

Discussion

In this section we investigate the theoretical and managerial contributions of our work as well as its limitations and possible avenues for future research.

Theoretical and Managerial Contributions

The knowledge management literature has emphasized the importance of knowledge sharing and knowledge creation for organizational success (Holten et al., 2016). Hence, non-sharing behaviors influenced by perceived supervision can lead to potential negative outcomes for organizations. In this context, the role of supervisors is critical as they have sufficient influence to drive employees to share or hide their knowledge (Kim et al., 2015; Khalid et al., 2018). The literature is abundant about negative effects that different facets of supervision may have individually on knowledge sharing within organizations; it is much scarcer about the possible positive effects of each of these facets, or of supervision considered as a holistic, multi-faceted construct, to discourage knowledge hiding behaviors. Few works in the literature have empirically investigated the relationship between supervision and a reduction in knowledge-hiding behavior. The first theoretical contribution of our study is the modeling of positively perceived supervision through its different facets (fair, supportive, and non-abusive), the development of the corresponding index to assess it through the adaptation of existing scales, and showing that the higher this index, the less employees are inclined to hide knowledge. Our work allows three different general supervision-related constructs/dimensions, previously studied independently and to our knowledge never together, to be indexed by a single quantity assessing positively perceived supervision, which provides higher criterion-related validity than its dimensions (Edwards, 2001). We argue that the aggregate construct of positively perceived supervision, and resulting index, is theoretically meaningful and parsimonious as an overall abstraction of the three highlighted dimensions.

Although several researchers have suggested that supervisors can have a destructive effect on knowledge-sharing behaviors (Kim et al., 2015; Lee et al., 2018), there is still a lack of research into supervisor behaviors that may discourage employee knowledge-hiding behaviors. The present paper starts to address this gap: in our work, we found that positively perceived supervision has a negative influence on employee knowledge hiding. This finding suggests that the way supervision is perceived by employees as a significant role to play in organizational knowledge management towards reducing knowledge hiding.

Table 1

Loadings and cross-loadings

Loadings and cross-loadings

-> See the list of tables

Table 2

Construct validity and discriminant validity – results of the reflective construct assessments

Construct validity and discriminant validity – results of the reflective construct assessments

-> See the list of tables

Table 3

Path estimates – measurement models – explanatory variables

Path estimates – measurement models – explanatory variables

*** Significant at p ≤ 0.001 ** Significant at p ≤ 0.01 * Significant at p ≤0.05

ns Not Significant

-> See the list of tables

Table 4

Structural paths

Structural paths

*** Significant at p ≤ 0.001 ** Significant at p ≤ 0.01 * Significant at p ≤0.05

ns Not Significant

-> See the list of tables

A further contribution of our study is the investigation of the role of organizational support and co-worker support on knowledge-hiding behaviors. Although researchers have previously found that organizational support and co-worker support can suppress the effects of abusive supervision and encourage people to share knowledge (e.g., Anand and Dalmasso, 2019; Kim et al., 2016; Lee et al., 2017;), surprisingly, we found that, when confronted with positively perceived supervision in the investigated nomological network, organizational support tends to encourage knowledge-hiding behavior, whereas co-worker support tends to reduce it. When organizational support, embedded in implemented rules and norms, is perceived as significant, this could create a sense of strong psychological ownership among employees, and thus encourage them to hide knowledge to defend their territory (Peng, 2013). Knowledge hiding can reduce innovation in organizations and become a threat to the organizational innovation process (Khalid et al., 2018). However, it has been shown that knowledge hiding may be reduced if supervisors adopt transformational leadership behaviors (Ladan et al., 2017). As a consequence of our findings, we would propose that organizations should perhaps be advised to consider ensuring that supervisors are trained to develop strong interpersonal relationships with employees, which can help to facilitate knowledge-sharing activities. In firms, beyond implemented rules and norms, supervisors should be made aware of the significance of their role and of the importance of the global perception of their multi-faceted supervisory role to discourage knowledge-hiding behaviors. Even though, perception of supervision may be considered subjective, it does help to understand how this perception is built. As a consequence, some managers could perhaps rethink their supervision styles, which in turn might contribute to improved collaboration and increased productivity of their firm.

Finally, our results contribute to international management research. Although knowledge sharing is important in every organization, our findings are grounded and validated in a cross-cultural data set that includes American, Indian, and French respondents as well as respondents from various other nationalities, all of them working in international firms. According to Boddewyn (2004), international management requires not only unidirectional movement of international borders, but also two-directional learning encountered by managers outside their home environments. Indeed, the quality and value of relationships between employees and their supervisor is a significant feature of most workplaces. However, cultural values play an important role in understanding how supervisors behave with co-workers, and how it may affect employees’ response to various aspects of their work environment (Kernan et al., 2011).

Therefore, managers / supervisors involved in managing and supervising subordinates, both physically and electronically, across various countries and cultural contexts, should pay specific attention as to how their behavior may be perceived through the lenses of different cultural values and norms as it will impact employee knowledge hiding behavior. Furthermore, and as cultural values are enacted differently at individual level (Debus et al., 2019), and coworker support appears important to discourage knowledge hiding, this social support should be nurtured by managers, more particularly in multinational firms.

Limitations and Future Research

This study has some limitations that point to new avenues for future research. We had to apply the snowball sampling method because, early in the data collection process, we discovered the reluctance of employees to disclose their perceptions of their supervisors (sensitive population: Goodman, 2011). We also purposely targeted knowledge intensive firms in the present study. Our number of participants did not allow us include control variables and effect multi-group comparisons. To extend our work, future research could focus on other specific types of organizations or groups (e.g., manufacturing, online firms, small firms, start-ups, blue-collar workers, etc.), and use larger samples if possible. One could for instance investigate whether a different non-probability sampling method (e.g., quota sampling) and a higher number of respondents would lead to the same cut-off thresholds for hypotheses 3a and 3b (p values = 0.031 and 0.05 respectively), reduce them, or perhaps lead to the verification of our hypotheses 3a and 3b, which was not the case in this study. One could also conduct multi-group analyses based on age, seniority, and / or years of service within a firm. Also, exploratory qualitative studies could provide a deeper understanding of investigated phenomena and future research may, therefore, benefit from being conducted using this type of approach to investigate further the causality link in the relationships examined in our study.

Furthermore, characteristics such as self-serving leadership (Peng et al., 2018) and personality antecedents, such as machiavellianism, narcissism, and psychopathy, have been found to impact knowledge-hiding behavior (Pan et al., 2018). Therefore, another possible avenue for further investigation is whether the behavior of supervisors is primarily based on their own individual characteristics or if it is partly due to work situations. Our study included respondents of various nationalities, but we did not have a sufficient number of respondents to perform a multi-group comparison. Some of our findings could differ from one cultural context to the next since culture, norms and policies have been identified as important predictors of KS behavior (Wang and Noe, 2010). Future studies could therefore investigate and compare different cultural contexts.

Conclusion

Hiding knowledge in organizations may be considered as unethical and unhealthy, leading to a reduction in employee innovation and a decrease in organizational performance (Hernaus et al., 2019; Serenko and Bontis, 2016). With the aim of understanding individuals’ knowledge-hiding behaviors in organizations, this empirical study investigated positively perceived supervision, modeled as a multi-dimensional construct with three different dimensions: supportive (work–family balance), fair, and non-abusive. The resulting index is found to negatively influence employees’ knowledge-hiding behaviors in relation to their supervisors and co-workers. Furthermore, we found that co-worker support reduced the propensity of employees to hide knowledge from their co-workers, whereas, surprisingly, organizational support increased their propensity to hide knowledge from their supervisor and co-workers.