An Empirical Study on Utilizing Pre-grant Publications in Patent Classification Analysis

Patent classification analyses are usually conducted using issued patents. Issued patents however suffer lengthy examination and the derived analytic results reflect R&D activities occurring considerable time in the past. The only option for an analyst to reduce such observational time delay is to use the so-called pre-grant publications (PGPubs) that are open to public 18 months after patent applications are filed. The PGPubs and their corresponding issued patents are both assigned classification symbols. If the two sets of symbols are very different, using patent classification analysis on PGPubs to observe R&D activities is dubious. This study therefore compares the United States Patent Classification (USPC) symbols assigned to about 235,000 pairs of U.S. utility patents issued in 2012 and their PGPubs in three ways, each corresponding to an approach of a conventional patent classification analysis: (1) considering only the class codes of the main classification symbols; (2) considering only the main classification symbols; and (3) considering both main and auxiliary classification symbols. The study finds that only the class codes of the PGPub main classification symbols are reliable enough for patent classification analysis as there are about 78% of the PGPubs have identical class codes as their corresponding issued patents. Conference Topic Patent analysis