Influencing factors of knowledge collaboration effects in knowledge alliances

ABSTRACT

Knowledge collaboration effects (KCE) can improve the efficiency of knowledge flow and knowledge sharing, and it is also an important way to generate value-added knowledge. Based on knowledge alliances (KAs) as the research object and exploratory and confirmatory analysis as the research method, this paper constructs the influencing factor system of KCE in KAs and judges on what degree of influence the factors have on KCE. The research result shows KCE in KAs are influenced by willingness to cooperate, learning abilities, knowledge attributes and knowledge activities. Some relevant suggestions are presented based on the study conclusion.

KEYWORDS:

• Knowledge alliances
• knowledge collaboration
• collaboration effects
• influencing factors

1. Introduction

Living in a knowledge economy era, business success of new product development is coming to rely more and more on knowledge (Pertusa-Ortega, Zaragoza-Sáez, & Claver-Cortés, 2010). A widely accepted definition is that “Knowledge management is an emerging set of principles that govern organisational and business process design, as well as specific processes, applications, and technologies that help knowledge workers dramatically leverage their creativity and ability to deliver business value” (Gurteen, 1998). Allee (1997) and Davenport and Pruzak (2000) think knowledge is professional intellect including know-what, know-how, know-why, and self-motivated creativity, or experience, concepts, values, beliefs and ways of working that can be shared and communicated.

Knowledge collaboration is defined broadly as the sharing, transfer, accumulation, transformation, and co-creation of knowledge involving individual acts of offering knowledge to others as well as adding to, recombining, modifying, and integrating knowledge that others have contributed (Faraj, Jarvenpaa, & Majchrzak, 2011). Knowledge collaboration is constantly affected by internal and external environment. When the internal and external control parameters reach their threshold values, the collaborative relationship among elements replaces the competitive relationship, transferring the system from disorder to order. In this way, collaboration effects are achieved (Cheng, Gu, & Quan, 2017; Imperial, Prentice, & Brudney, 2018). As shown in Table 1, although the research field of knowledge collaboration is still in its initial development stage, various views have been put forward by different scholars.

Table 1. Different views of knowledge collaboration

View	Resource
Theory of process
Knowledge collaboration is a multi-dimensional dynamic process to realise knowledge innovation	Mckelvey, Almb, and RiccaboniI (2003); Tong (2011)
Knowledge collaboration is a process to conduct interactive learning and knowledge innovation in the form of two-way dynamic flow	Lundvall and Bengt (1992); Cock (2002); Zou and Wang (2016)
Knowledge collaboration is a developing process from partial collaboration to integral collaboration	Jonathon and Sara (2005); Meng and Zhang (2016)
Theory of mobility
Knowledge collaboration is a mobile activity involving different individuals in a clustered environment	Li (2009)
Knowledge collaboration is an innovative activity exploiting knowledge resources	Salavisa, Sousa, and Fontes (2012)
Theory of synergy
Knowledge collaboration has value-added effects, of which the overall benefits outperform the sum of benefits of independent parts	Wu and Gu (2012)

In general, views about knowledge collaboration can be classified into three types: theory of process considers knowledge collaboration as a process or procedure to achieve innovations of knowledge, the process is constantly progressing, emphasising how knowledge collaboration comes into being. Theory of mobility stresses the mobility and flexibility in the knowledge collaborative activity, it is changeable according to environment and resources. Theory of synergy, commonly agreed by most scholars, reveals that knowledge collaboration helps form a “1 + 1 > 2” result, generating surplus value by synergistic action.

Knowledge alliance is a network organisation which links knowledge with knowledge, and gathers organisation members together in network space to achieve knowledge innovation as the goal (Lavie, Haunschild, & Khanna, 2012). KA is formed by knowledge flow among different organisations with the purpose of achieving knowledge innovation. The extensive connection among organisations makes an ideal KA, in which knowledge gets value-added (Inkpen, 1998). With the advent of knowledge economic era, knowledge has become a critical part of economic resources and an important source of competitive advantages (Peter, 1988). In order to improve core competence, enterprises keep seeking for knowledge resources by forming strategic alliance and partnership with universities, research institutes, suppliers, customers and even competitors (Anklam, 2005; Makino & Delios, 1996; Nieves, Quintana, & Osorio, 2016; Yayavaram, Srivastava, & Sarkar, 2018). In this way, KAs achieve knowledge sharing, knowledge creation and knowledge advantage and finally benefit from KCE (Gu, Guo, & Li, 2003). At present, existing research results about KAs mainly focus on the model of knowledge, knowledge flow, knowledge sharing, knowledge creation as well as cooperative management, conflict management and risk management of KAs, which lack relevant study about knowledge collaboration in KAs. Wu and Gu (2008) put forward a knowledge collaboration mechanism including opportunity identification mechanism and process mechanism. Shi, Gu, and Wu (2011) construct the principal-agent model in KAs through calculation of collaborative coefficients. Zhang (2014) defines the order parameters of knowledge collaboration from width, depth and strength to construct knowledge collaboration mechanism model in KAs. Hu, Gu, and Cheng (2015) point out that KCE of KAs could be formed by subject coordination, coordination mechanism and knowledge collaboration. Cheng et al. (2017, 2018) respectively study knowledge collaboration pattern and knowledge collaboration management.

From the above, we can extract that, for an enterprise in a knowledge economic era, a major method to grasp core competence is to facilitate its knowledge collaboration with other knowledge alliance partners. However, it is not known which elements may have an effect on KCE. Based on previous study, this research attempts to explore the influencing factor model, which represents a new approach to knowledge management aimed at achieving knowledge collaboration. This paper describes the formation mechanism of knowledge collaboration from the next four aspects: willingness to cooperate, learning abilities, knowledge attributes and knowledge activities. Willingness to cooperate can be divided into three types: mutual trust, mutual benefit and mutual dependence, based on which organisation members seek for cooperative partners and reach common strategy targets. Meanwhile, good learning ability of members promotes the organisation to better absorb knowledge, which means further motivation for knowledge collaboration is provided by learning autonomy, absorptive capacity and applying ability. And then, affected by different attributes of knowledge, KAs realise knowledge division, knowledge flow, and knowledge sharing and knowledge creation. The whole process makes an ideal knowledge collaboration mechanism, which is shown in Figure 1.

Figure 1. Research framework

Under the mechanism of knowledge collaboration, individuals integrate the internal and external resources of an organisation to make the overall benefits outperform the sum of benefits of independent parts, just like “1 + 1 > 2”. Through collaboration, more benefit could be generated to enhance the core competence, therefore, knowledge collaboration creates much greater value than the “1 + 1 = 2” physical cooperation (Chen & Chen, 2002; Wu & Gu, 2012). KCE is affected by various elements and it is of vital importance to explore these elements to expand KCE.

The composition of this paper is as follows. Section 2 introduces the theory, extant literature and puts forward hypotheses accordingly. Section 3 uses questionnaire method to collect data. Section 4 carries out empirical research and discusses the results through a series of methods. The last section includes the concluding comments. The framework of the whole research design is shown in Figure 2.

Figure 2. Knowledge collaboration mechanism

2. Theory and hypotheses

2.1. Willingness to cooperate

Willingness of individuals to cooperate is the premise of knowledge collaboration. A strong willingness to collaborate promotes knowledge to transfer, making it easier for individuals to cooperate and improve the efficiency of knowledge collaboration (Triandis, 1995). The relationship strength of KAs affects knowledge collaboration via the level of trust, interaction frequency, interdependence of resource and degree of asset specialisation.

Firstly, knowledge sharing is a process along with risk. Since high degree of trust could reduce transaction cost and decrease the opportunism behaviour, it is necessary to consider mutual trust in a governance mechanism to promote knowledge sharing and knowledge innovation (Han, Lee, Beyerlein, & Kolb, 2018; Klijn, Edelenbos, & Steijn, 2010). Organisations, which have built mutual trust, are honest to each other. These organisations have full faith in each other’s resource and motivation, which means cooperative relationship, is more transparent. In such a way, knowledge stickiness is weakened and efficiency of knowledge transfer is enhanced.

Secondly, under the principle of mutual benefit, an equal and coordinating relation is built between the two sides. Organisations accept each other and cooperate with each other (Triandis, 1995). Based on mutual benefit, knowledge transfer could conquer obstacles existing in the traditional model and achieve perfection. In a reciprocal environment, organisations form a stable cooperative relationship on the basis of resource sharing and complementary advantages, which makes knowledge flow more flexible.

Thirdly, if a certain kind of resource is not accessible in factor market resulting from it special features, then allying could be considered as a main approach to acquiring this resource (Gulati, 1995). Through KAs, organisations build an interdependent cooperation system to provide conditions for knowledge flow. This leads us to hypothesise the following:

H1: The willingness to cooperate of individuals has a positive effect on knowledge collaboration of KAs.

2.2. Learning abilities

As a result of the unpredictable social environment, learning ability of each individual has become a crucial factor that influences organisation development. Individuals who are good at identifying learning opportunities could absorb new knowledge more positively. These individuals transfer received knowledge to personal working experience. This suggests that improving learning ability of individuals helps build a platform for organisation members to share and study knowledge. In this way, knowledge is kept in a more free-flowing state, which makes individuals clearer about the characteristics and external environment of an organisation. Individuals keep studying, summarising and exploring the future direction. These steps make up for a virtuous cycle to ensure the sustainable development of the organisation.

As a part of learning ability, learning autonomy shows initiative and self-awareness in leaning behaviour. In the process of self-learning, individuals firstly make clear learning objectives. Then, by searching and selecting relevant knowledge in certain fields, individuals select the most valuable parts to learn and form their own understanding. Organisation members do not only absorb professional knowledge but also cultivate the capacity for solving problems and adapting to the environment (Dickinson, 1995). The improved comprehensive quality of individuals helps to realise talent-post matching and enhances the competitive ability of the organisation.

Absorptive capacity of knowledge can be considered from potential and actual perspectives. Potential absorptive capacity reflects knowledge acquisition, and actual absorptive capacity reflects knowledge transfer (Zahra & George, 2002). More specifically, as we mentioned at the start that knowledge mainly consists of professional skills and experience of doing things, individuals get to learn the basic definition of knowledge by absorbing these skills and experiences in a potential way. However, knowledge at this stage only remains at the theoretical level, which needs to be further developed into personal working experience and techniques. This further development is determined by individuals’ actual absorptive capacity such as the ability to accept new things (Patterson & Ambrosini, 2015). Therefore, we can infer that absorptive capacity, especially actual absorptive capacity, is a critical factor to promote the circulation of external knowledge within an organisation. Thus absorptive capacity is also crucial to promote knowledge collaboration.

An important step of knowledge collaboration is to apply the knowledge into practical business. The acquired knowledge cannot be fully utilised only when it gets recombined and turns into targeted methods and techniques. Great applying ability of knowledge makes individuals flexibly put knowledge into practical use (Ayadi, Brandwayn, Clark & Ayadi, Brandwayn, Clark, & Wagner, 2005, 2005). By combing new and old knowledge, individuals understand knowledge by analogy, and give full play to knowledge. These arguments prompt us to hypothesise the following:

H2: The learning ability of individuals has a positive effect on knowledge collaboration of KAs.

2.3. Knowledge attributes

Knowledge has been defined in various ways, it is not homogenous and nondistinctive. Knowledge can be classified in terms of different characteristics and the basic distinction from each other makes knowledge own attributes in certain situations (Spender, 1996). For example, as the object of knowledge collaboration, the availability and suitability of knowledge are usually emphasised. In addition to transferability determined by explicit and implicit knowledge, knowledge attributes also include embeddedness due to specific environment and complementarity caused by knowledge gap. These attributes exert influence on knowledge collaboration from diverse perspectives.

Embeddedness is also known as dependency. This attribute implies that knowledge is generated based on different forming conditions and the sphere of influence. Knowledge plays its role in a specific work domain and social environment, while becoming inapplicable under other circumstances (Balland, Belso-Martínez, & Morrison, 2015). High level of embeddedness shows that knowledge is sticky and dependent, and only possessed by a sole organisation, which makes it hard for other organisations to imitate. This phenomenon is detrimental to knowledge collaboration as the cost of knowledge transfer is inevitably increased. Based on Granovetter (1985)’s view, we deduct that knowledge, whether explicit or implicit, all has embeddedness and knowledge embeddedness impedes the process of knowledge collaboration.

Complementarity of knowledge shows the relationship between different parts of knowledge that is mutually explained or reinforced. Complementarity exists in two forms: The complementarity of different parts of the same type of knowledge along with time experience, and the complementarity of different types of knowledge along with spatial experience (Wang, 1997). In knowledge collaboration of KAs, knowledge complementarity is usually spatial. On the basis of different knowledge base and knowledge requirement, organisations reach common sense and share knowledge with each other to make up for their own knowledge disadvantages as well as meeting the knowledge requirement of the other side. In this way, knowledge system is perfected. Hence, organisations should establish KAs with partners who have complementary knowledge to keep stable inflow and outflow of knowledge.

According to the degree of difficulty of being transferred, knowledge is divided into explicit and tacit knowledge. Explicit knowledge has a higher transferability while tacit knowledge has a lower transferability. Explicit knowledge can be transferred from one side to the other side, thus can be easily absorbed. Tacit knowledge, if is to be shared and transferred, must be externalised to easily-understood form. In order to speed up the process of knowledge transfer, organisations should not only seek for explicit knowledge but also actively promote the transformation from tacit knowledge to explicit knowledge (Ambrosini & Bowman, 2001). Ways of developing and managing tacit knowledge include holding training courses and sharing sessions. These standard procedures help realise the uniform distribution of knowledge and improve the study effect of the whole organisation (Smith, 2001; Cavusgil, Calantone, & Zhao, 2003). Accordingly, the following hypothesis is developed:

H3: Knowledge attributes have a positive effect on knowledge collaboration of KAs.

2.4. Knowledge activities

In order to attain knowledge collaboration, knowledge activities among KAs members are also of great importance in addition to willingness to cooperate, learning abilities and knowledge attributes. More specifically, knowledge activities include knowledge division, knowledge flow, and knowledge sharing and knowledge creation. As knowledge is divided into various fields in the charge of different members or organisations, knowledge gap thereby comes into being. What is more, this phenomenon will continue pushing knowledge flow, speeding up knowledge sharing and promoting knowledge creation.

Firstly, knowledge division means that each member of KAs is focused on one specific knowledge field and only needs to master that kind of knowledge. Knowledge division forms knowledge branch that spreads all walks of life. Organisations on each branch have knowledge of different areas and stocks (Becker & Murphy, 1992). By exchanging knowledge with each other, knowledge in the KAs tends to be evenly distributed to maximise the benefit.

Secondly, knowledge flow refers to the diffusion and transformation of knowledge among different individuals participating in innovation activities (Gu, Li, & Wang, 2006). Knowledge flow can be divided into knowledge flow among organisation members and knowledge flow across organisations according to the scope of flow. As a result of knowledge division and knowledge gap, it is necessary for knowledge alliance members to carry out communication activities to promote knowledge transformation from one side to the other. In this way, knowledge system can be improved so as to promote the two-way flow of knowledge, and realise the innovation and reorganisation of knowledge (Mu, Peng, & Love, 2008).

Thirdly, knowledge sharing implies the dissemination of knowledge between individuals within and across the organisation through various channels in order to achieve knowledge innovation and generate benefit. Knowledge sharing is closely connected to knowledge flow: Knowledge flow promotes the diffusion of knowledge in different organisations, which is the premise of knowledge sharing. Knowledge sharing is the in-depth study of the required knowledge, which is the subsequent development of knowledge flow (Zhuge, 2002). Although most scholars believe knowledge sharing is a determinant to increase knowledge creation and innovation (Gunu & Ajayi, 2015; Pittino, Barroso Martínez, Chirico, & Sanguino Galván, 2018), there have also been empirical studies in which a direct relationship between knowledge sharing and innovation was not found (Darroch & McNaughton, 2002). But above all, it has been evidenced that sharing of knowledge achieves collaboration effects and collective learning, therefore having an indirect impact on innovative results (Chen, Huang, & Hsiao, 2010). Additionally, there are generally two views on the factors of knowledge sharing: Some scholars hold the view that individual’s psychological characteristics promote the sharing process of knowledge (Cabrera, Collins, & Salgado, 2006). The others believe organisational interventions encourage knowledge-sharing activities more (Almeida & Kogut, 1999; Thompson, Gentner, & Lowenstein, 2000).

Lastly, knowledge creation is the ultimate embodiment and the advanced stage of collaboration effects, which will only occur when knowledge collaboration develops to a certain stage. Knowledge creation is shown by mutual transformation of explicit knowledge and tacit knowledge. Japanese scholar Nonaka proposes the SECI model, which divides the process of knowledge creation and dissemination into four stages: socialisation, externalisation, combination and internalisation. To be more specific, socialisation means that the members of an organisation gain experience through personal experience, imitation and perception. Externalisation manifests itself in receiving the instruction of others and conceptualising tacit knowledge. Combination is the reintegration of explicit knowledge into new knowledge. Internalisation means that individuals will reapply the new knowledge, after reorganisation, to work practice (Nonaka & Takeuchi, 1995). The whole process is cyclic and spiralling, in which knowledge is divided, flowing and shared, and finally, knowledge creation and knowledge collaboration are realised. These arguments lead us to hypothesise:

H4: Knowledge activities have a positive effect on knowledge collaboration of KAs.

3. Methods

3.1. Data sources

The aim of this research is to explore the factors influencing KCE in KAs. The problem we explore is about knowledge, a set of skills or experience of every individual, which is abstract and cannot be easily qualified like the financial target or performance of an enterprise. Accordingly, the questionnaire method was chosen the most appropriate vehicle (Krosnick, 2018). Questionnaire analysis is unified, highly standardised, with strict testing of validity and reliability of every question item. Moreover, apart from quantitative data, questionnaire method is also expected to gain information marked by attributes, qualities, and attitudes. The recipients of this study are faculty and staff working for enterprises, colleges and research institutes in Sichuan Province in China. Therefore, social media appears to be a convenient and viable way for data collection. Questionnaires were distributed through social media such as email and websites. Besides, the questionnaire data were recorded in an online form, which helps the data processing work and safeguard the recipients’ personal information.

We received 268 questionnaires initially. To enhance the validity of the samples, we set a pre-test before the formal question item to investigate basic information of the respondents, including the property of their affiliated organisations and departments, the actual operation term of organisations, etc. Respondents removed include the following cases: (1) their affiliated organisations had not enough exchanges with others. For example, we set some question items such as the number of suppliers, the number of major clients, duration of business with suppliers, duration of business with major clients, if their answers for these questions were 1–5 suppliers or major clients, or business duration less than half a year, these samples would be rejected. We set these pre-test question items to guarantee the samples we chose were from large-scale organisations, having frequent business contact and communication with others, thus the establishment of knowledge alliance and the emergence of KCE would be more likely to occur in these organisations. (2) The chosen items were ambivalent. For example, if a respondent’s answer for the question the operation period was less than 2 years but the answer for another question the length of knowledge exchanges with suppliers was more than 2 years, this sample would be deleted as it had logical contradiction. (3) We distributed some questionnaires in online form through Sojump (a professional online platform in China providing questionnaire survey services). We used Sojump to eliminate some samples with total time of answering less than 60 s, as the abnormally brief time to answer the questionnaire may imply hasty attitudes, therefore leading to invalid data. Similarly, some questionnaire with abnormal answering distributions, for example, all the answers were completely agree or completely disagree, or not finishing the whole questionnaire, were kicked off. With this method, 221 out of 268 questionnaires were screened with the effective rate of 82.5%, showing a strong representation.

3.2. Data features

In terms of the organisational nature, enterprises account for 40.83% of the total, while universities and colleges account for 17.43%. Scientific research institutions account for 11.47%. Others are government departments, financial institutions and advisory bodies. In terms of the life of organisations, 52.29% of the samples have been established for over 10 years, accounting for the majority of research objects. From the department of respondents, 40.37% of respondents belong to R&D department, 27.52% are from management department, and 16.51% and 15.6% are from functional department and logistics department, respectively. From the length of knowledge communication with customers, 55.05% of the organisations have established knowledge exchange with customers for more than 2 years, while the smallest percentage was 11.01% with less than half a year.

3.3. Survey design

The questionnaire can be divided into three parts: (1) related concepts (2) basic information of the samples (3) main content of this questionnaire. The first part introduces the theoretical and practical purpose of this research, explains relevant concepts to the recipients, including the definitions of KAs and KCE. The second and third parts are in Tables 2 and 3. The basic information of samples helps conduct descriptive analysis to know the basic distribution of samples and kick off some unqualified questionnaires.

Table 2. Questionnaire to collect information of recipients

Question	Answer item
The property of your affiliated organisation is:	Enterprise Research institution Financial institution	College Governmental department Others
The number of staff in your organisation is:	Under 300 501-1000	301-500 Over 1000
Your affiliated department is:	Management department Logistics department	R&D department Functional department (procurement, production and sales)
The operation period of your organisation is:	Less than 3 years 6–10 years	4–5 years Over 10 years
The number of suppliers contact business contact with us is:	1–5 11–20	6–10 Over 20
The number of major clients contact business contact with us is:	1–5 11–20	6–10 Over 20
Business duration with suppliers:	Less than half a year Less than 2 years	Less than 1 year Over 2 years
Business duration with major clients:	Less than half a year Less than 2 years	Less than 1 year Over 2 years

Table 3. Questionnaire to evaluate influencing factors on KCE in KAs

Please answer these questions in forms of scoring on the right side			Completely diagree→completely agree
Variable	Unit	Question	1	2	3	4	5
Willingness to cooperate	Level of trust	A1 Organisations that trust each other have more frequent knowledge exchanges
(Wu & Gu, 2008)		A2 Organisations that trust each other have deeper knowledge exchanges
	Mutual benefit	A3 Organisations obtain what they need and benefit together through knowledge exchanges
	Mutual dependence	A4 Knowledge exchanges between cooperative partners are more frequent
		A5 Knowledge exchanges between cooperative partners are deeper
Learning abilities (Luo et al., 2017)	Learning autonomy	B1 Organisation members should take the initiative to acquire knowledge
		B2 Organisation members should form the habit of self-learning
	Absorptive capacity	B3 Organisation members should understand the basic concepts of the knowledge gained
		B4 Organisation members should have personal understanding of the knowledge gained
	Applying ability	B5 Organisation members should be skilled in knowledge and techniques
		B6 Organisation members should apply their knowledge to practical operation
		B7 Organisation members should innovate in the knowledge acquired
Knowledge attributes (Xiao et al., 2009; Xue & Sun, 2012)	Embeddedness	C1 Knowledge acquired should conform to the orientation of enterprise development
		C2 Knowledge acquired should conform to specific economic and political environment
	Complementarity	C3 Knowledge between organisations is complementary
		C4 Knowledge exchanges between organisation members can complement each other and progress together
	Transferability	C5 Knowledge acquired should be easily understood
		C6 Knowledge acquired should be easily transformed into words or data
Knowledge activities (Gu et al., 2006)	Knowledge division	D1 Unevenly distributed knowledge between organisations is fit for knowledge exchanges
		D2 Knowledge of different domains between organisations is fit for knowledge exchanges
	Knowledge flow	D3 Organisations should conduct frequent exchanges of knowledge, technology and talents
		D4 Organisations should make the advantage of social networks to exchange knowledge, technology and talents
	Knowledge sharing	D5 Organisations should share knowledge through exchanges and learning
	Knowledge creation	D6 Knowledge is transformed into a new kind of knowledge through communication and sharing
KCE (Hu et al., 2015)		E1 Organisation members gain more skills and expertise
		E2 Benefits of organisations are increased
		E3 Influence of organisations is enhanced

The main content of questionnaire is designed in 5-point Likert scale. The 24 items used to assess KCE were based on the discussion of Gu et al. (2006), Wu and Gu (2008), Xiao, Gu, and Peng (2009), Xue and Sun (2012) and Luo, Wei, and Gu (2017), including willingness to cooperate, learning abilities, knowledge attributes and knowledge activities. Another three items to describe KCE are based on the research of Hu et al. (2015). Every question item is described in a declarative sentence. Recipients need to judge every item and expressed their degree of recognition by scoring. The scores, ranging from 1 to 5, show different opinions from completely disagree, disagree, to neutral, agree and completely agree.

We divided the whole questionnaire into five parts accordingly, and each part includes several question items to measure the corresponding hypothesised variables. Specifically, from our statements of hypotheses, the independent variables are willingness to cooperate, learning abilities, knowledge attributes and knowledge activities, the dependent variable is KCE. Therefore, we set a series of questions for multi-dimensional measurement of the five variables (four independent variables and one dependent variable). These questions, as we have mentioned in Table 3, are put forward based on extant study of scholars. Previous researchers have provided evidence for us to set standardised questions for each variable, so all the question items are of great persuasion. Similarly, the recipients we pick are also representative. In 3.1 part, we introduced how we carried out the sample selection work, trying to avoid the error caused by the insufficient sample representativeness. What the recipients need to do is to judge and score for each question. Moreover, the hypotheses we put forward were all positive, therefore, to confirm the hypotheses, the next analysis work for us is to evidence the respective question score of the four independent variables all have positive correlations with the dependent variables’ question score.

4. Results and discussion

4.1. Reliability and validity of questionnaire

This study uses the SPSS software to conduct reliability and validity analysis of the questionnaire. Reliability mainly reflects whether the data have internal consistency. The internal consistency coefficients of the reliability of the unity are weighted sums of the item scores. We use Cronbach’s Alpha coefficient to test the reliability. Cronbach’s Alpha coefficient is the average value of the split-half reliability coefficient obtained from all possible division methods of the scale items, generally ranging from 0 to 1. Therefore, large Cronbach’s Alpha coefficient implies greater reliability. After testing, the Cronbach’s Alpha value of 27 variables reaches 0.949, exceeding the 0.8 threshold, which shows that reliability of sample data is very high, having very good stability and reliability.

Validity refers to the degree to which instrument truly measure what it is intended to measure. We carried out the Bartlett’s test of sphericity and calculate the Kaiser–Meyer–Olkin (KMO) statistic. Bartlett’s test of sphericity is a method to test the degree of correlation between variables, KMO statistic is used to compare simple and partial correlation coefficients between variables. The two criteria are necessary pre-tests to check the validity of the questionnaire model, and also the pre-tests for factor analysis in the next part – if the test result shows great correlation between variables then factor analysis can be supported by these variables. In this study, Bartlett’s test of sphericity was significant (p < 0.0001), KMO statistic is 0.937, higher than the threshold value 0.8, suggesting that individual measures of sampling adequacy were generally high, and strong correlation exists between variables.

4.2. Factor analysis

Table 3 represents the extraction of principal component factors. Based on the above, this study adopts principal component analysis (PCA) to test the validity. PCA was developed in order to exam whether the hypothesised variables could be identified empirically, and for testing the reliability and validity of the measures. Firstly, factors with characteristic root greater than one are extracted, and then the common factors are rotated orthogonally through the maximum variance method. The rotation converges after seven iterations, and four principal component factors are extracted, attributing to 61.711% of the total variance. This percentage means the four extracted factors could cumulatively explain 61.711% of the whole questionnaire information, which shows that each variable has great structural validity. Specifically, the loadings of variables are all greater than 0.4, with the highest loading reaching 0.729. Together, the cumulative explanation and the loading of variables indicate that all the common factors have the ability to well explain the independent variables (Table 4).

As it can be seen from Table 5, the first common factor includes items B1 – B7, which reflects the ability of organisational members to learn, understand and apply knowledge independently and we name it as learning ability. The second common factor includes the items A1 – A5, which reflects the need for mutual trust and cooperation between organisations as well as the degree of interdependence, and reflects the willingness of organisations to cooperate. Therefore, this factor can be named as willingness to cooperate. The third common factor includes the items C1 – C6, which describes the nature of knowledge from three aspects including the degree of embeddedness between knowledge and environment, the complementarity between knowledge, and the degree to which knowledge can be taught. So in this paper, this factor is named as knowledge attributes. The fourth common factor includes items D1 – D6, reflecting different stages of knowledge activities from four aspects of knowledge division, flow, sharing and creation, which is named as knowledge activities.

Table 4. Total variance explained

	Initial Eigenvalues			Extraction Sums of Squared Loadings			Rotation Sums of Squared Loadings
Component	Total	% of Variance	Cumula-tive%	Total	% of Variance	Cumula-tive%	Total	% of Variance	Cumula-tive%
1	11.161	46.502	46.502	11.161	46.502	46.502	4.085	17.022	17.022
2	1.520	6.333	52.836	1.520	6.333	52.836	3.668	15.281	32.303
3	1.126	4.691	57.527	1.126	4.691	57.527	3.562	14.843	47.146
4	1.004	4.184	61.711	1.004	4.184	61.711	3.495	14.565	61.711

Table 5. Rotated component matrix

Qu	Component				Qu	Component
	1	2	3	4		1	2	3	4
B1	0.720				C1			0.715
B2	0.707				C6	.		0.701
B5	0.654				C2			0.674
B6	0.639				C3			0.601
B3	0.559				C5			0.559
B4	0.551				C4			0.444
B7	0.495	.			D3				0.684
A5		0.729			D6				0.657
A2		0.715			D5				0.656
A1		0.710			D2				0.583
A4		0.687			D1				0.556
A3		0.563			D4				0.545

4.3. Correlation analysis

Through factor analysis, this paper makes a preliminary summary of the influencing factors of KCE in KAs and verifies the hypothesis. However, correlation analysis and regression analysis are needed to further understand the close degree and causal relationship between the influencing factors. Therefore, firstly, correlation analysis is carried out on the four factors that have been extracted. As shown in Table 6, Pearson correlation coefficients of willingness to cooperate, learning ability, knowledge attributes and knowledge activities are all greater than 0.4, which indicates that there is a positive correlation between the factors and the KCE in KAs, and the correlation is good, indicating that there is a close relationship between the factors.

Table 6. Correlations

	Willingness to cooperate	Learning abilities	Knowledge attributes	Knowledge activities
KCE	0.446**	0.468**	0.420**	0.406**

*Significance level at 10%.

**Significance level at 5%.

***Significance level at 1%.

4.4. Regression analysis

The two variables in the correlation analysis are not divided into independent variables and dependent variables, which cannot further reflect the causal relationship between various influencing factors and knowledge collaboration. Therefore, in order to explore the specific effect of various influencing factors on the KCE, this study uses the stepwise regression method to further analyse the direction of the variable relationship. As shown in Table 7, the four variables enter the regression equation in turn, with tolerance of 0.655, 0.765, 0.680, and 0.671 separately, which are all greater than 0.6, indicating that the tolerance of each variable is good, and there is no obvious linear coincidence among the four variables. The B value indicates the coefficients of the independent variables of each influencing factor in the regression equation, but since the dimensions and range of values of each independent variable are different, the B value does not reflect the influence magnitude of each independent variable on the dependent variable, so reference is made to the standardised coefficient Beta. Among them, the Beta value of willingness to cooperate is 0.381, which is the largest, followed by knowledge attributes, learning abilities and knowledge activities. It can be seen that the order of the influence degree of each influencing factor on the KCE is willingness to cooperate, knowledge attributes, learning abilities and knowledge activities, and the model of KCE influence factor of KAs can be obtained as shown in Figure 3.

Table 7. Regression

	Unstandardised Coefficients		Standardised Coefficients			Collinearity statistics
Model	B	Standard error	Beta	t	P	Tolerance	VIF
（Constant）	1.979	0.256		7.721	0.000
Knowledge attributes	0.540	0.053	0.362	10.279	0.000	0.655	1.526
Willingness to cooperate	0.614	0.053	0.381	11.696	0.000	0.765	1.308
Learning abilities	0.478	0.057	0.290	8.4045	0.000	0.680	1.471
Knowledge activities	0.253	0.049	0.181	5.195	0.000	0.671	1.489

Figure 3. The factor model of KCE in KAs

4.5. Confirmatory analysis

We use structural equation model (SEM) to carry out confirmatory analysis. SEM is a method to establish, estimate and test causal models, containing both observable explicit variables and potential variables that cannot be directly observed. SEM has diverse functions such as multiple regression, path analysis, factor analysis, covariance analysis and other methods, and can clearly analyse the effect of single index on the whole and the relationship between single indexes.

We use AMOS software to verify the parameters of the model. CMIN/DF (χ2/df) = 2.33 < 3, which shows that the degree of fitting is good. NFI = 0.944, IFI = 0.906, CFI = 904, GFI = 0.921, which are all greater than 0.9, and meet the standard. Therefore, this model can be used to analyse the relationship between various influencing factors and their influence on knowledge collaboration. The standardised coefficients are shown in Figure 4.

Figure 4. SEM model

From the analysis results, it can be seen that under the condition of complete standardisation, the covariance parameters between the four influencing factors are 0.92, 0.89, 0.90, 0.77, 0.86, 0.78, which indicates that the degree of correlation between the willingness to cooperate, learning abilities, knowledge attributes and knowledge activities is relatively high. At the same time, the covariance parameters of the 27 variable indexes for each influencing factor is greater than 0.7 (except Q20, which is 0.65), indicating that the model has strong explanatory power and good fitting performance.

As is shown in Table 8, willingness to cooperate, learning abilities, knowledge attributes and knowledge activities have positive influence on the KCE. The standardised coefficients are all around 0.5, C.R. >1.96, P < 0.005, showing that these four factors can promote knowledge collaboration, and the effect is significant. Therefore, H1, H2, H3 and H4 are true.

Table 8. Hypotheses and validation

Hypothesis	Standardised coefficient	C.R.	Conclusion
H1: The willingness to cooperate of individuals has a positive effect on knowledge collaboration of KAs.	0.55	5.281	True
H2: The learning ability of individuals has a positive effect on knowledge collaboration of KAs.	0.50	5.293	True
H3: Knowledge attributes have a positive effect on knowledge collaboration of KAs.	0.51	5.336	True
H4：Knowledge activities have a positive effect on knowledge collaboration of KAs.	0.46	4.140	True

5. Implications for practitioners

Willingness to cooperate of individuals has a positive influence on the KCE. Organisations should take the initiative to establish knowledge exchange with the outside and maintain a stable frequency of knowledge sharing. At the same time, organisations should maintain mutual trust and enhance the transparency of cooperation.

The ability of organisational members to learn, understand, absorb and integrate knowledge is positively related to KCE. Organisations should encourage members to be positive to learn in forms of training, holding lectures and other activities to help members better understand and use the knowledge they have acquired, and encourage members to innovate and integrate knowledge so as to further develop the practical value of knowledge.

The embeddedness, complementarity and transferability of knowledge have influence on knowledge collaboration to a certain extent, which requires organisations to select knowledge that meets their own requirements in a targeted way and establish KAs with external partners. Meanwhile, through internal communication and training, the degree of explicitness of knowledge can be improved so that knowledge can be transmitted and understood more easily.

6. Conclusions

Knowledge collaboration is an advanced stage of knowledge management, which could promote knowledge sharing, expanding the knowledge base and producing knowledge value-added effect. Based on the research results of scholars at home and abroad, using questionnaire method and combining theoretical analysis with empirical analysis, this paper divides the influencing factors of KCE in KAs into four aspects: willingness to cooperate, learning ability, knowledge attributes and knowledge activities, and each aspect is set with several sub-indexes. The empirical results show the increase of the four aspects above helps improve KCE in KAs.

This study has developed a deeper insight into the research on KCE. For example: (1) Hu et al. (2015) used questionnaire method to point out KCE of KAs could be formed by subject coordination, coordination mechanism and knowledge collaboration. Also through questionnaire method, this study adds factor analysis to classify the items in the questionnaire and proves the items indeed reflect different perspectives of influencing factors of KCE in the hypotheses part. Factor analysis, correlation analysis and regression analysis, as a whole, provide evidence suggesting that the questionnaire design has great reliability. Besides, in the last part of discussion, we use SEM method to verify the degree of fitting of the model and clearly reflect different interrelations of the indexes. (2) When it comes to influencing factors of KCE, a large number of prior studies focus on the human factors (Han et al., 2018; Klijn et al., 2010; Patterson & Ambrosini, 2015; Triandis, 1995). Apart from the human behaviours, this study especially considers the attributes and activities of knowledge as the influencing factors. In fact, in the forming process of KCE in KAs, we could regard the human behaviour as the subject (i.e. willingness to cooperate, learning abilities) and knowledge as the object (i.e. knowledge attributes, knowledge activities). Combining these two aspects improves the index system.

In spite of all the findings in line with our hypotheses, limitations of our work include the fact that the existed KCE in KAs is limited to the four influencing factors above without considering other possible factors. Additionally, this study was limited to data from Chinese organisations. Because the KCE may depend on organisational cultures, future research needs a more representative sample of the general organisations for each country. Moreover, further amendment to enhance the general applicability of the mechanism may be required.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the National Social Science Foundation of China under Grant [CIA140187], the Southwest Petroleum University Humanities Special Foundation under Grant [2017RW006] and the Southwest Petroleum University Youth Science and Technology Innovation Foundation under Grant [2018CXTD14].