An Empirical Analysis of Patent Citation Relevance and Applicant Strategy

Abstract

Patent examination should ensure that only novel and nonobvious technologies are patented. This evaluation requires comparing the invention to technologies described in public documents—called “prior art.” Examiners and applicants have obligations to cite known prior art that is material to whether the patent is issued. Beyond documenting examination, citations are used as metrics in a significant body of research. The importance of citations as a predictive metric rests on the assumption that they provide evidence of continued development in the relevant field. Research indicates that some citations are, however, made for reasons beyond technological similarity. This undermines the notion that citations show continued growth of a technology. We analyze this assumption—and correct for inaccuracies—by employing similarity metrics to characterize the “relatedness” of technologies described in two patent documents (i.e., citing and cited references). To this end, we use a “Jaccard Index” to quantify textual similarity—and thus technological relatedness—of two documents. Using this method, we empirically analyze strategic behaviors in patent law that were previously only theoretically described in the literature. For example, some patent applicants “bury” relevant references—submitting many irrelevant references and a few relevant ones to hinder review of the important ones. Our Jaccard Index analysis is the first to empirically evaluate whether this practice benefits the applicant. Moreover, we improve upon patent value and grant rate analyses and demonstrate that citation relevance has a significant impact above and beyond a count of citations made.