knowledge source characteristics

Analyzing innovative activities and the interplay between science and technology often relies on patent statistics as valuable proxy measures. These statistics offer rich data sources for studying innovation and technical advancement (Callaert et al., 2006; Chandra & Dong, 2018; \u00c9rdi et al., 2013; Fleming & Sorenson, 2001; Hall et al., 2005; Jaffe & Trajtenberg, 1999; P. C. Lee, 2019, 2019; Liaw et al., 2014; No & Park, 2010; Pavitt, 1985; Pezzoni et al., 2022; Trajtenberg, 1990). This study utilizes patent data from the United States Patent and Trademark Office (USPTO), which is widely recognized as reflecting global technological developments (\u00c9rdi et al., 2013; Hall et al., 2005). Moreover, technological activities inherently involve both backward references to previous inventions and subsequent forward citations by other inventions. Unlike backward citations, which are fixed at the time of patent approval, forward citations can vary over time. Therefore, to minimize truncation bias and enhance the objectivity of data distribution, the computation of forward citations in this study is configured to span a minimum of five years post-approval. To enable a comprehensive five-year post-approval citation analysis for all patents, USPTO patents granted between 1976 and 2014 have been used as observation base, with forward citation calculations including data from 1976 to 2019.To mitigate concerns of data skewness and potential prediction bias, a meticulous sampling process was undertaken from a base of 4,971,086 patents. Each group contained roughly 100,000 patents, and this sampling process was repeated three times to ensure consistency and uniformity in the results. This procedure produced 112,364 patents from the original STC pool (3,514,034 patents) and uncovered an overlap of 89,166 patents with the initial FTC pool (94,692 patents). Post stratified sampling, the total patent count reduced to 204,853 from the original 4,971,086.