Exploring the impact of external collaboration on firm growth capability: the mediating roles of R&D efforts

Sample

Strategy& and PwC released the 2018 Global Innovation 1000 Study, reporting the world’s top-ranking 1000 listed companies with the highest R&D expenditure. Together, they accounted for 40% of total global R&D expenditure. This study shows that R&D expenditures increase worldwide, but most notably in China, rising 34.4 percent in the year. As an emerging country with rapid development in recent years, China has made great contributions to the development of the global new economy.

We collected and analyzed data of Chinese innovative enterprises ranked in the above study for three reasons. First, these enterprises are in line with the sustainable development strategy advocated by the Chinese government. Most of them are leaders in their respective industries. Second, these enterprises attach great importance to R&D and have made great efforts in R&D. Third, under the support of sufficient resources and attraction of sustainable development, most of them have implemented external collaboration and developed cooperative alliances. Moreover, the information disclosure system involving listed companies provides us with more convenience for data collection.

We followed the following procedures to select our sample. First, we picked out 175 Chinese companies in the 2018 Global Innovation 1000 Study. Second, 76 Chinese companies listed in Hong Kong, Taiwan, and United States were excluded because of the difficulty of data collection. Third, even another 5 companies in mainland China with missing data were also excluded. Finally, 94 Chinese innovative companies listed on Shenzhen and Shanghai Stock Exchanges (A share) were selected as our sample for data analysis.

Table 1 displays the characteristics of our sample. The mainly involving industries included the information technology industry (26.6%), capital goods industry (28.7%), advanced materials industry (17.0%), automobiles and components industry (11.7%), consumer durables and apparel industry (7.4%), and retailing and media industry (2.1%), healthcare industry (4.3%) and energy industry (2.1%). The companies comprised 60.6% state-owned firms and 39.4% non-state-owned firms. Among them, 17.0% hired less than 10000 employees, 42.6% between 10001 and 30000, 19.1% between 30001 and 50000, and the other 21.3% hired more than 50000 employees. In addition, the sample aged 6–10 years accounted for 7.4%, 11–15 years for 12.8%, 16–20 years for 40.4%, 21–25 years for 31.9%, and finally more than 25 years for 7.4%.

Data

The data collection proceeded in two stages. In the first stage, we built composite scales for measuring external collaboration and process dimensions of firm growth capability. We marked the scales based on a content analysis of the company’s disclosed information. In the second stage, the data of other variables were collected from China Stock Market & Accounting Research Database (CSMAR Database).

Following the increasing use of panelists in research (e.g., Zott and Amit, 2007), we set up a panel composed of 3 members, including 1 professor and 2 doctoral students. First, the professor carefully selected our panelists from his research team, requiring that the selected team members should have a good understanding of the firm’s external collaboration and growth capability. After choosing the most qualified candidates, the professor claimed the two selected members (doctoral students) to carefully read the information, announcements, and documents of total sample companies, get familiar with the details of external collaboration and growth capability, and develop measurement scales by following inductive logic. Next, the professor made further training for the two doctoral students, who were authorized as expert raters in data collection and analysis. In addition, the raters were provided with written guidelines on the proper way to address survey items. The underlying materials for data collection include annual financial reports, corporate social responsibility reports, investment analysts’ reports, company news, company websites, and other company announcements between 2016 and 2020. It took every rater about six months from October 2020 to April 2021 to collect data on external collaboration and firm growth capability’s process dimension. To reduce the influence of common method bias, the process of data collection was divided into two stages: the one is to collect the data of independent and control variables, and the other is to collect the data of mediating and dependent variables. The interval between the two is one month. The lack of readily available data obliged us to draw on primary sources of data and constructed a unique, manually collected dataset. The method also prevented us from collecting time-series data. Finally, we evaluated the consistency by conducting a pairwise comparison of two raters’ scores, yielding a Pearson correlation coefficient of 0.951 (p < 0.01). For the different scores, the two raters discussed with each other under the guidance of the professor and finally reached a consensus.

The data of other variables were drawn from CSMAR Database, which is a research-oriented accurate database in the economic and financial field compiled by Shenzhen CSMAR Data Technology Co., Ltd. The database combines the actual national conditions of China and draws on the professional standards of authoritative databases such as CRSP, COMPUSTAT, TAQ, and THOMSON. The period of data collection was fixed as the year 2018. To control the influence of extreme values on research results, the data collected from CSMAR were winsorized.

Measures

Independent variables

In line with previous studies (Garcia Martinez et al., 2019), we distinguished three types of external collaboration and set them as independent variables. That is vertical, horizontal, and competitor collaboration. Among them, vertical collaboration (VC) was measured by the total score of collaboration with suppliers and customers; horizontal collaboration (HC) was measured by the total score of collaboration with governments, universities and educational institutions, consultancy firms, venture capital investment firms, trade fairs and exhibitions, and others; and competitor collaboration (CC) was measured by the score of collaboration with competitors. The score of collaboration with each type of collaborator was coded as a binary variable which was measured by 1 and 0. Among them, 1 represents that collaboration widely and deeply happens and 0 represents the reverse side.

Dependent variables

Considering several studies have explored the relationship between external collaboration and innovation performance (Findik and Beyhan, 2015; Lu and Yu, 2020), we need to define and measure firm growth capability (FGC), the dependent variable of our study, from a new perspective that differs from the prior studies. We realized that firm growth capability is a much more comprehensive construct than innovation performance, and so we tried to measure it from both financial and process dimensions according to the theory of balanced scorecard. We carried out a data pre-processing for the financial and process ones into [0,1] interval and measured the firm growth capability by the mean value of their scores. First, the financial dimension of firm growth capability was measured by four indexes, i.e., debt-paying ability, operating ability, profit ability and development ability. Among them, the debt-paying ability was assessed by the current ratio (current assets/current liabilities); operating ability was evaluated by inventory turnover (operating costs/average inventory); profitability was measured by operating profit ratio (operating profit/operating income); and development ability was evaluated by the operating income growth rate (current year’s operating income/last year’s operating income −1). Through principal component analysis, we calculated the weighted score of the financial dimension. Second, we independently developed a scale for measuring the process dimension of firm growth capability. Four items, i.e. “the enterprise has established a global R&D center and created a global intelligent R&D platform”, “the enterprise has advanced intelligent manufacturing system”, “the enterprise has diversified, advanced, and intelligent sales system for offline and online sales”, and “the enterprise has advanced, convenient and intelligent service system for dealing with pre-sale, sale and after-sale issues”, were developed to respectively measure research and development ability, manufacturing ability, marketing ability, and serviceability. Obviously, the firm growth capability measured in our study includes but also goes beyond the connotation of innovation performance that was concerned by predecessors (Findik and Beyhan, 2015; Lu and Yu, 2020). Considering the difficulty to obtain an objective score for the measures, we deemed the use of perceptual coding of our raters. The involved items were quantified on a five-point scale. After coding, the scores of total items were aggregated and averaged as the final score of the process dimension.

Mediating variables

R&D intensity and R&D human capital constitute R&D effort. Among them, R&D intensity (RI) was measured by the ratio of a firm’s R&D expenditure to operating income (Kim and Lee, 2016), and R&D human capital (RHC) was measured by the percentage of highly skilled R&D workers (researchers and technicians) (Teixeira and Tavares-Lehmann, 2014).

Control variables

Learning from the previous study (Yu and Yan, 2021), we designed four control variables as the alternative explanations for firm growth. First, we set ownership as a dummy variable that controls for potential variations between state-owned enterprises (coded as 1) and private-owned, foreign-owned, or other types of enterprises (coded as 0). Second, the age was assessed by subtracting the year of firm establishment from the year in which the survey was conducted. Third, the size was measured by the natural logarithm of the employee scale. Finally, the industry as a dummy variable was coded as 1 when a firm belongs to the industry of advanced materials, consumer discretionary, and healthcare and energy. Otherwise, it was coded as 0.

Statistical technique

Following the suggestion of Hox (1994), we tested direct effects by hierarchical regression analysis, using SPSS 24 software. Following the recommendation of Preacher and Hayes (2008), We tested mediating effects by bias-corrected bootstrapping procedure, using PROCESS v. 3.3.

Model development

Drawing on hierarchical regression analysis, we constructed the following models to test our hypotheses.

First, to measure the impacts of external collaboration on firm growth capability that were proposed by H1(a, b, c), we constructed the models M1-M5. Among them, model M1 tests the effect of control variables, models M2-M4 respectively test the effects of vertical collaboration, horizontal collaboration, and competitor collaboration, and model M5 tests their combined effect.

Second, we developed the models M6-M15 to test the effects of external collaboration on R&D efforts that were proposed by H2(a, b, c). Among them, M6 and M11 were set for measuring the effects of control variables on RI and RHC respectively; the models M7-M9 measure the independent effects of VC, HC, and CC on RI, and similarly the models M12-M14 measure their independent effects on RHC; and the models M10 and M15 test their combined effects on RI and RHC respectively.

Next, we built the models M16-M18 to examine the impacts of R&D effort on firm growth capability that were proposed by H3. Among them, the models M16 and M17 measured the independent effects of RI and RHC respectively, and in M18 we measured their combined effect.

Finally, we developed the models M19-M21 to measure the mediating effects of RI and RHC on the relationship between external collaboration (VC, HC, and CC respectively) and FGC that were proposed by H4(a, b, c).

where Controls indicate control variables, including ownership, age, size (Ln), and industry, μ indicates the random disturbance, and i denotes the number of samples.