Starting with a theoretical discussion on R&D collaboration, technological boundaries, and innovation performance, the importance of a firm's collaboration and technological boundaries for its technological innovation performance was empirically analyzed using patent data. The results of the analysis can be summarized as follows: a company's cooperation has positive effects on both exploitation and exploration. Exploitation has a positive impact on a firm's quantitative innovation performance, while exploration has negative effects on a firm's quantitative innovation performance.
The relationship between a firm's research activities and a firm's qualitative innovation performance manifests as an inverted U-shape. On the other hand, a firm's exploitation activities have a U-shaped relationship with the firm's qualitative innovation performance. We investigated the effect of a firm's amount of collaboration experiences on its technological innovation performance.
They found that non-competitive collaborations have direct positive effects on a firm's innovation performance, and both sources of collaboration have positive moderating effects on internal R&D efforts on innovation performance within firms.
Technology boundary spanning and a firm’s innovation performance
They found that all types of persistent collaborations have positive effects on the company's innovation, while only recently concluded collaborations with universities or research institutes significantly improve the company's innovation. Although collaboration has many positive effects, such as reducing risks and costs, it can create several potential risks. Cooperating partners may engage in opportunistic behavior such as cheating and misrepresentation (Das & Teng, 1998).
Hypothesis
The firm’s collaboration and technology boundary spanning
The firm’s collaboration experience and innovation performance
Both studies used binary variables to understand collaboration variables, but this study uses the variables that reflect the number of collaborations. According to Allen's (1983) argument, firms can jointly invent new technology by sharing their knowledge with their competitors. In other words, the collaboration enables companies to obtain a larger amount of patents that could represent the company's quantitative innovation performance.
However, Cowan and Jonard (2003) indicated that strongly collaborative firms may suffer negative effects due to repetition. In highly collaborative networks, exchanged knowledge between neighborhoods is similar and can lead to redundancy in inventions. As a result, we can expect that a firm's collaboration may affect the average number of citations received because the competitors already possess similar technology on their own.
Technology boundary spanning and a firm’s innovation performance
We used quantitative innovation performance as the result of innovation and qualitative innovation performance as the process of innovation. Based on the above explanation, we can expect that exploitation will have a positive effect on quantitative innovation performance, which is the current result, while exploration will have a negative effect. Furthermore, we hypothesize that exploitation has a negative effect on qualitative innovation performance due to the firm's ability to adapt to environmental changes.
Hypothesis 4c: The relationship between a firm's research activities and its qualitative innovation performance is curved (inverted U-shaped).
METHODS
Data
In this paper, we use two indicators: the number of patents and the number of citations received. Because the distribution of patent value is highly skewed, we cannot determine the innovation rate of firms by the number of patents alone. This paper attempted to test the effect of several factors not only on quantitative innovation performance measured as the number of patents, but also on qualitative innovation performance measured as the average number of patent citations the firm received.
In the case of the business layer, we used collaboration to express the relationship between companies. The nodes include all companies involved in the printer industry each year. The size of a node was determined by the number of patents owned by the company.
The edge between firm F1 and firm F2 represents the collaboration between firm F1 and firm F2, and the edge weight represents the number of collaborations between them. In the case of the technology class layer, we used class coupling to present the relationship between classes. The edge weight refers to the number of patents that include class C1 and class C2 together.
A high level of class coupling between class C1 and class C2 implies a high possibility of combining C1 and C2 when the firm in the printing industry is looking for a new invention. The interactions between two layers provide the distribution of technology classes in which the firm invests. In the case of a patent that has more than two technology classes, the edge weight made by that patent is one divided by the number of technology classes for each technology class.
Variables
To measure a firm's collaboration experience, we used the degree of centrality in firms' collaboration network. To measure a firm's collaborative status, we used the ratio of the collaborated patents to the total number of patents. We counted the company's patents that were the result of collaboration and divided this figure by the total number of patents the company held.
To measure a firm's exploitation and exploration, we used the boundary between a firm and a patent class in a multilayer network that included a firm and a patent class. We defined exploitation and exploration as the firm's patenting rate on the existing patent class and any new patent class. To check the market situation and concentration of the company, the average market size of the class in which the company invested (MARKET_POSITIONING) and the number of classes in which the company invested (CLASS_IN) were used.
First, we designed two models to confirm the effect of a firm's cooperation on exploitation and exploration, respectively. Second, we designed models for the firm's quantitative innovation performance, which was measured by the number of patents the firm owned. In Model 3, we included the network density of the firm's collaboration network (DENSITY), the average market size of the technology the firm invested in (MARKET_POSITIONING) and the number of technology classes the firm invested in (CLASS_IN).
Because the degree centrality in the collaboration network may be directly related to the number of patents, we used the ratio of the collaborated patents to the total number of patents to test the relationship between the collaboration and the firm's quantitative innovation performance. To capture the effect on the company's qualitative innovation performance, we used the average number of citations over five years. Therefore, the regressions on the firm's qualitative innovation performance used degree centrality in the collaboration network.
RESULTS
- Descriptive Statistics and Correlations
- A firm’s collaborations and technology boundary spanning
- The Firm’s Quantitative Innovation Performance
- The Firm’s Qualitative Innovation Performance
- Robustness check
The models in Table 5 report the results for the panel regressions of firm quantitative innovation performance using fixed effects (FE). The results in Table 5 show that the network density of the firm collaboration network (DENSITY) has a negative impact on the firm's quantitative innovation performance. The average market size of the technology in which the firm invests (MARKET_POSITIONING) also has a negative effect on the firm's quantitative innovation performance.
Otherwise, the number of technology classes the firm invests in (CLASS_IN) positively influences the firm's quantitative innovation performance. Models in Table 6 report the results for the panel regressions for firm qualitative innovation performance using fixed effects (FE). The results in Model 7 show that the network density of the business collaboration network (DENSITY) negatively affects the firm's qualitative innovation performance.
The average market size of the technology in which the company invests (MARKET_POSITIONING) has a positive influence on the qualitative innovation performance of the company. Moreover, the number of technology classes the company invests in (CLASS_IN) also positively influences the company's qualitative innovation performance. In Model 8, we introduce degree centrality in the business collaboration network (DEGREE) as an independent variable.
However, the results show that the square term of exploitation (EXPLOITATION^2) has a positive effect on the company's qualitative innovation performance, which means that the relationship between the exploitation activities and the company's qualitative innovation performance is curvilinear (U-shape) . The square term of exploration (EXPLORATION^2) has a negative effect on the firm's qualitative innovation performance, making the relationship between exploration activities and the firm's qualitative innovation performance curvilinear (inverted U-shape). There was no change in the results for the control variables or in the degree of centrality in the company's collaboration network (DEGREE).
Contrary to the results in Model 10 and Model 11, the network density coefficient of the firm's collaboration network (DENSITY) had a positive value. The results showed that only the squared term of exploration had a statistically significant effect on firm quality innovation, which supports hypothesis 4c.
CONCLUSION AND DISCUSSION
Our results show that degree centrality in the collaboration network has a positive effect on the average number of citations received. Therefore, our results imply that excessive collaboration could lead to a decrease in invention quality but an increase in the number of inventions. At the same time, we propose a new way to measure exploitation and exploration using a multi-layer network and confirm its effects on the firm's innovation performance.
Our results show that exploitation has a positive effect on a firm's quantitative innovation performance, while exploration has a negative effect on it. In addition, our results show that the squared exploration time has a negative effect on the qualitative innovation performance of the firm. Thus, we confirm the inverted U-shaped relationship between research activities and the qualitative innovation performance of the company.
Furthermore, an appropriate level of exploration can be beneficial to firms by providing higher qualitative innovation performance. We acknowledge that this research, limited to a sample of 10 years in the printer industry, may not be fully generalizable. Furthermore, although we used a multilayer network only to measure exploitation and exploration, it can be developed in many different directions.
From a “firm to class” perspective, the degree of firm centrality, related to the boundary between the layers, can be interpreted in terms of the technology class the firm is working on. The betweenness centrality of the firm can be attributed to the role of the hub in merging the technologies. Rather, in “class to firm” terms, the degree to which class is central can be attributed to the number of firms competing in the technology class.
Effects of research timing on innovation: The value of being out of sync with rivals. Technology transfer as technological learning: A source of competitive advantage for firms with limited R&D resources.
