Abstract:      With the intensification of market competition and the increasing complexity of technology, the R&D cost of firms and the market risk faced by firms are increasing, and it is more difficult for firms to deal with market competition by relying only on internal knowledge resources. When there are resource and capacity constraints within a firm, it can obtain the technical knowledge which is lacking in the firm through inter-firm knowledge sharing in the R&D cooperation. In theory, firms should search for partners as far as possible in order to maximize innovation efficiency. However, due to the limited capacity and information asymmetry of firms, firms can only search and select partners within a limited range according to their resource attributes of firms, among which the knowledge attributes of firms and the knowledge relations between firms have increasingly become the value coordinates for firms to search for partners.
   Although there are abundant studies on enterprise knowledge attributes (such as diversity, complexity, etc.) and knowledge relations (relatedness), the existing research mainly focuses on one or two aspects of the knowledge attributes of partners, the knowledge attributes of focal firms and the knowledge relations between cooperative firms. Few literatures have integrated the three aspects into the same analytical framework to comprehensively study the influence of R&D cooperation on the innovation performance of focal firms. In fact, knowledge attributes of firms themselves and knowledge relatedness between cooperative firms do not independently influence the innovation performance of firms. On the one hand, the firms′ own knowledge attributes are the foundations for firms to establish knowledge relations and determine the content and mode of knowledge sharing between firms. On the other hand, different knowledge attributes of firms influence each other and jointly determine the innovation performance of firms.
    Therefore, based on the knowledge-based view and organizational learning theory, this paper establishes a comprehensive analysis framework on the impact of knowledge attributes and inter-firm knowledge relations on innovation performance of firms, and studies the impact of knowledge diversity of partners, knowledge relatedness between partners and focal firms on focal firms′ innovation performance, as well as the moderating effect of focal firms′ knowledge complexity. In order to better study how knowledge attributes and knowledge relations affect firms′ innovation performance in R&D cooperation, this paper regards R&D cooperation as the process of knowledge sharing between firms, and further divides knowledge sharing into two stages: knowledge access and knowledge integration. From the two stages of knowledge activity process, this paper deeply analyses how knowledge attributes and knowledge relations affect firms′ innovation performance in R&D cooperation. 
     In the knowledge access stage of knowledge sharing, firms contact a wide range of knowledge domains of partners, identify and acquire the required knowledge, and conduct shallow learning. At this time, the knowledge diversity of partners, that is, the degree of diversity of knowledge domains of partners as a whole, has a role in promoting learning. In the stage of knowledge integration, firms integrate some acquired domain knowledge through deep learning, so as increasing the diversity of their own knowledge domains. In the process of knowledge internalization, the knowledge relatedness between focal firms and partners will affect the absorption efficiency of distinctive knowledge and the risk degree of internal unique knowledge leakage, thus affecting the effect of knowledge integration. At the same time, the firms′ own technological environment determines the payoff from knowledge recombination. The knowledge complexity of the focal firm, as a representative of the recombination potential of the internal knowledge and the complexity of the knowledge architecture, has a certain impact on the knowledge acquisition intention in knowledge access and the efficiency and effectiveness of the knowledge internalization in knowledge integration. Then, using 45 listed pharmaceutical companies in 2007-2016 as samples, an empirical test is conducted by negative binomial regression model to test hypotheses. Finally, this paper puts forward conclusions and suggestions for firms to improve their innovation performance by improving the R&D partner selection mechanism.
    The results show that: (1) The knowledge diversity of partners has a positive impact on innovation performance of focus firms. In the knowledge access stage of knowledge sharing, the higher the degree of knowledge diversity of partners, the wider the scope of knowledge contact and acquisition by firms, which provides more possible space for focal firms to integrate knowledge, and the higher the innovation performance of focal firms. (2) Knowledge relatedness between partners and focal firms has an inverted U-shaped effect on innovation performance of focal firms. In the stage of knowledge integration, with the increase in knowledge relatedness between firms, the homogeneity of knowledge base increases, which is more conducive to communication between both sides, improving the absorptive capacity of new knowledge and increasing the efficiency of knowledge integration. When the degree of knowledge relatedness is too high, heterogeneous knowledge between cooperating firms becomes less. It is difficult for focal firms to absorb new knowledge to stimulate innovation, and the risk of leakage of their own unique knowledge combination increases, resulting in a downward trend in innovation performance. (3) Knowledge complexity of focal firms negatively moderates the relationship between knowledge diversity and innovation performance. When the degree of knowledge complexity is high, the positive effect of partners′ knowledge diversity on innovation performance of firms is significantly weakened. (4) Knowledge complexity positively moderates the inverted U-shaped relationship between knowledge relatedness and innovation performance of focal firms. When the degree of focal firms′ knowledge complexity is high, the cooperation with the firms with higher knowledge relatedness is easier to form the innovation output.
    The theoretical significance of this paper contains the following two points. First, an analytical framework is established to analyze the impact of knowledge attributes and inter-firm knowledge relations on the innovation performance of focal firms, which provides a theoretical basis for the study of partner selection mechanism based on the firms′ knowledge characteristics. Second, this paper expands the research on knowledge sharing and firm innovation, combines static knowledge characteristics with dynamic knowledge activity process, and clarifies the mechanism of knowledge attributes and inter-firm knowledge relations influencing innovation performance of firms in R&D cooperation. The practical significance of this paper lies in the establishment of an analytical framework for the impact of knowledge attributes and inter-firm knowledge relations on firms′ innovation performance. The empirical conclusions based on this framework have important reference value for firms to improve the knowledge-based partner selection mechanism.

Key words:  R&D cooperation, knowledge attribute, knowledge relation, knowledge sharing, innovation performance