Number of Backward Patent Citations as a Quantitative Indicator of Prior Art (Japanese)

Patent citations have been widely utilized to measure inter-organizational knowledge flows. Measurement of knowledge flow is important in understanding how "open innovation" influences inventive activities through such means as personnel movement, venture capital, technology licensing, and R&D cooperation with universities. However, it is known that patent citations contain noise in measuring knowledge spillover. Moreover, there remains another question -- whether or not inventors' citations evenly reflect inventors' recognition of prior knowledge within and across organizational boundaries. Since some citations come from inventors' own organizations, prior patents within the organization might show up as citations more often than those outside of organizational boundaries. If that is the case, biased measurement of cumulative innovations through citations should be considered when we try to evaluate knowledge diffusion within hierarchies and in the market. With this motivation in mind, this study matches backward citations with the RIETI Inventor Survey. In particular, survey responses concerning the dependence on prior patents and the location of the prior patents (whether or not prior patents come from the same firm) are analyzed in combination with backward citations. First, backward citations are confirmed to be consistent with the survey responses in the sense that the ratio of internal citations over outside citations is positively and significantly correlated with the survey responses, suggesting that prior patents existed internally. However, further analyses using binary choice models show asymmetric results regarding internal and outside citations. Only when there are more citations inside inventors' firm boundaries, does the probability of a positive response to the survey question, "the invention was substantially dependent on prior patents" increase. In contrast, the number of outside citations is not a significant factor. Only the number of internal citations seems correlated with the probability of inventors' perception of the prior patents that they were dependent upon. The results hold not only with the Japan Patent Office (JPO) inventor citations and with U.S. citations overall, but also with JPO examiner citations. Since JPO examiners have little incentive to focus only on prior arts inside the inventors' organization, the "locality" of inventors' (revealed) awareness of prior patents creates a new puzzle. We can at least conclude that, compared with the number of backward self-citations, the number of non-self citations has a different character as a proxy variable for inventors' perceptions. A follow-up inventor survey is planned to scrutinize how citations within firm boundaries differ from those across firm boundaries.