The results of the structural equation model are shown in Figure 2 and support Hypotheses 1a, 1b, 3a, 3b, 4a, 4b, and 5.
SN has a positive effect on both AKS (β=0.17; p <0.1) and SNKS (β=0.145; p<0.1). Individual and group perceptions of the advantages of fostering and enabling the existing social network while promoting the quality of interactions impacts the behavioral intention of knowledge sharing (Razmerita et al., 2016) between peers in healthcare research centers (Park & Gabbard, 2018). Given the natures of the items addressed in the construct, the frequency, quality, and common perception of approachability have a positive effect on both the individual cognitive evaluation and predisposition to share knowledge. Conversely, a positive norm behind group membership and personal networking between members has a positive effect towards the cognitive construction of collective subjective norms that support knowledge sharing, as the literature shows (Lang, 2004; Chow & Chang, 2008). Accordingly, the additional perception of approachability and trustworthiness behind ST that comes hand-in-hand with the SN is also found to have a positive effect with both the AKS (β=0.31; p<0.1) and SNK (β=0.37; p<0.1)
SG doesn’t have a significant effect on both AKS and SNK. Considering these results, Hypotheses 2a (p>0.1) and 2b (p>0.1) are not supported, questioning previous literature (Nahapaet & Goshal, 1998; Inkpen & Tsang, 2005). Nonetheless, the specific and social complexities of knowledge, especially in its tacit form, focus more on the frequency and reciprocity of communication to improve knowledge performance in academic research teams, where the individual goals and the social motivation can be addressed as a barrier to knowledge sharing (Yu et al., 2018). Multi-team systems with interconnected teams, often common in healthcare research teams, whose design can lead to countervailing forces when diffusing broad goals, can slow the collaboration between peers who become progressively more distant from those goals (National Research Council, 2015), which can be a significant reason for these findings.
Both AKS (β=0.524; p<0.001) and SNKS (β=0.533; p<0.001) have a significantly positive effect on KSI, which confirms the TRA principles of behavioral intention mediated by the empirical weights of both the individual’s attitude and the social norm present in the individual´s social environment (Ajzen, 1985; 1992). Thus, the higher the behavioral intention, the higher the behavior outcome (β=0.533; p<0.001).
To better understand the behavioral outcome of knowledge sharing, we apply a method, valuable for exploring and examining complex causality (Fiss, 2011), that relies on Boolean algebra principles when considering the configurations of causal conditions. fsQCA accepts a variety of conditions and paths to reach a given outcome (equifinality) and it explores causal asymmetry by addressing the necessary conditions for the presence and for the absence of the outcome (Fiss, 2011), offering a holistic approach to understanding the interaction and independencies between the antecedents and other variables that are further understood when not in isolation. (Short et al., 2008). The fsQCA uses the same variables as the SEM to enlarge the results from the quantitative analysis. Other demographic variables are considered to compose the configurations for KSB (García-Sanchez et al., 2017), enlarging the analysis.
Calibration transforms the data set into fuzzy-set membership scores. Following Ragin (2008), variables (conditions in fsQCA) are calibrated using the direct method into three anchors to allow classification of the conditions from full membership (1.00) to full non-membership (0.00) for each score, and threshold for the most ambiguity in the fuzzy set (0.5). The calibration technique rescales the condition data with the support of these anchors (Fiss, 2011). The transformation of the Likert scales data is done via a calculation of the average values of the construct for each latent variable (Woodside et al., 2011). To further expand the analysis, the KSB condition is also rescaled by considering the scores on two groups of items (tacit KSB and explicit KSB). Table 4 displays the statistics and calibrations of the causal conditions.
We develop the suggested inspection of both intermediate and parsimonious solutions (Fiss, 2011) and thus we identify the core conditions (present in both the parsimonious and intermediate solutions) and the peripheral ones (only present in the intermediate solution). Sufficiency analysis respects the threshold for raw consistency of the configurations (0.80) (Ragin, 2006). The coverage values presented in the solutions also respect the suggested range in the literature (0.25-0.90) (Ragin, 2008; Woodside & Zhang, 2013). Tables 5, 6, 7, and 8 report the intermediate solutions. The black circles (⚫) indicate the presence of a condition, and empty circles (⚪) indicate its absence. Larger circles indicate core conditions. Small ones, indicate peripheral conditions. Blank spaces indicate a condition that does not matter for the configuration.
Results show different configurations for the presence and absence of KSI (three for the presence and two for the absence of the outcome). AKS (atitu) is a core condition in all three configurations leading to KSI (int). But, SNKS (norm) is not present in any of these configurations. Regarding the discussed antecedents, only one configuration presents the presence of both ST (trust) and AKS (atitu) in order to reach KSI (int). The other antecedents, however, are either absent or do not contribute to the configuration (Table 5). Hence, Hypothesis 6a is supported.
Regarding the absence of KSI (~int), we find different core conditions in the solution. The results show that ST (trust) leads both to the presence and the absence of the outcome in different configurations, while there are other conditions that remain absent in both the presence and the absence of the outcome. Conversely, AKS’s absence (~atitu) accounts alone for one of the configurations as a core condition that leads to the absence of KSI (~int). Hence, Hypothesis 6b is supported.
Regarding the analysis for the configurations that lead to tacit KSB (tacit) and explicit KSB (explit), results show a difference in characteristics, such as work experience, seniority, and age in the configurations for the presence and absence of the outcomes (Tables 6 to 8). The absence of KSI (~int) in all configurations for the absence of KSB (~beh) (KSB, tacit KSB, and explicit KSB) confirms that using the TRA to address KSB is a valid approach (Ho et al., 2009). Additionally, the presence of young researchers (~age) also seems to inhibit knowledge sharing.
Although some configurations are common to the three KSB outcomes, some differences are addressed in the analysis. For example, more experienced and tenured male researchers with high KSI tend to share more knowledge in one configuration for overall KSB (beh) but not for explicit KSB (explit), where the absence of seniority seems to be a peripheral condition to explicit knowledge sharing (explit). Conversely, while the absence of seniority (~sen) is also present in some of the configurations that lead to explicit KSB (explit), its presence is relevant in different configurations that lead to the absence of explicit KSB (~explit). Therefore, seniority is a barrier to explicit knowledge sharing, which is consistent with García-Sanchez et al.´s (2017) findings of a negatively significant moderation of PhD seniority on the trust of the research team in KSB that affects explicit knowledge sharing. On the other hand, tacit KSB (tacit) is an outcome for both the presence and absence of the seniority in the organization. While young (~age) and “junior” (~sen) researchers in the organization share tacit knowledge between themselves, there is a gender difference. A configuration for the outcome tacit KSB (tacit) shows that young female researchers with less experience and seniority in the organization are more likely to share knowledge.
On the other hand, more experienced male researchers with less seniority in the organization are more prone to engage in tacit KSB. While the presence of KSI (int) is found in configurations for both KSB (beh) and explicit KSB (explit), it is also absent for tacit KSB (tacit). The degree of high abstraction and the lack of its operationalization, at a conscious level, found in the literature regarding tacit knowledge can justify its ever-constant flow in the social life of the organization (Nonaka & Takeuchi, 1995; Boisot, 2002; Choo, 2006; van den Berg, 2013) even when individuals seem to lack a behavioral intention to share it. The richness of alternative configurations that lead to the presence and absence of KSB, explicit KSB, and tacit KSB support Hypotheses 7a, 7b, 8a, 8b, 9a, and 9b.