臺大管理論叢 NTU Management Review VOL.29 NO.1

163 NTU Management Review Vol. 29 No. 1 Apr. 2019 To measure a firm’s technical capabilities, we collected patent data from the Webpat database. This database provides most of the information available on the introductory text of a granted patent. More importantly, it offers worldwide patent information, including patents granted by the patent and trade offices in Taiwan, United States, Europe Union, Japan, and China. Our dependent variable—firm performance—is measured by Tobin’s Q and the growth rate of sales. Our main independent variables—the two types of external networks—are measured by the number of external R&D contracts and supply chain contracts, respectively. To measure a firm’s technical capabilities, we counted the number of granted patents that a firm holds. To measure a firm’s manufacturing and marketing capabilities, we use the value-added to sales as a proxy. We measure the governance modes of external networks by counting the number of equity-based agreements among all external R&D and supply chain contracts. Since the dataset is in the panel format, we employ a random-effects GLS regression model to address concerns of unobserved heterogeneity (Greene, 2003). A fixed-effects analysis is also used and included to be a robustness check. 4. Results The results support all our hypotheses (H1a, H1b, H2a, H2b, H3a, and H3b, respectively). In order to verify the robustness of the results, we conducted several robustness checks. First, we adopted a fixed-effects regression model to estimate firm performance. According to the results, we found that the estimated main effects of the external networks and the two moderating effects remained at a level of significance. Second, we drew the interaction graphs (See Figures 2 to 9 from p.153-157) and followed Aiken,West, and Reno (1991) to conduct a series of simple slope analyses. These robustness checks supported the above-mentioned results.