Abstract: Based on the misclassification model, this paper evaluates grant decisions made by State Intellectual Property Office (SIPO). The sample bases on G class of foreign invention patents (foreign is defined as priority is filed outside of China) and filing date in SIPO is during 2010-2015. The empirical results show that our country’s patent examination did not show a significant "false grant" phenomenon. However, there is about 4.4% of patents, whole oversea family patents obtained lots of forward patent citations, were not granted by SIPO, which is defined as “false rejection”. The proportion of "false rejection" risen to 5.8% in software-related patent samples. Furthermore, based on recognition of the erroneous decision, this article discussed the relationship between patent and patent applicants’ characters and the probability of the false rejection.

摘要： 基于Hausman et al.,(1998) 错分模型，文章对2010年－2015年期间国家知识产权局受理的G大类发明专利申请的授权决定进行了误差检测。这里的授权误差的定义为，优先权在国外进行专利申请并获得较多后续专利引用的专利，其中国同族专利的申请未能获得授权，即“过度驳回”；亦或优先权在国外进行专利申请但未获得任何后续专利引用，其中国同族专利的申请却获得授权，即“过度授权”。基于此样本的实证结果显示，中国专利审查并未显示出显著的“过度授权”现象。但是，有大约4.4％的专利，其外国同族专利获得了高频次后续专利引用，并被中国的专利审查“过度驳回”。“过度驳回”的比例在软件相关专利样本中上升至5.8%。基于错分模型对过度驳回审查决定的识别，文章进一步检测了权利要求数量，技术应用广度，审查经验等因素对于中国专利审查过度驳回概率产生的影响，并提出了相应的政策建议。