Page 13 -
P. 13
means by which differences in patent quality were accessed, especially in large datasets
where in-depth qualitative evaluation of individual patents is not possible. Citations to
prior patents are submitted as part of a new patent application in order to establish “prior
art.” The applicant is required to disclose the prior art to the examiner and must then
establish how their invention goes beyond that prior art. The examiner often adds
citations they are aware of, such that half of the citations come, on average, from the
examiner [41]. While it is not incumbent upon the applicant or the examiner to be
exhaustive in their citation inclusion, important patents do on average receive more
citations [34]. Similar to publications [42], inventors (authors) often cite their own work
in part to show the evolution of knowledge but also, perhaps, to influence their citation
relevance [43, 44] or Eigenfactor [45], which are used as the basis of tenure and other
promotion decisions [46]. Therefore, for all analysis herein, patent technology value is
defined by citations from future patents less inventor self-citations, henceforth denoted
simply as citations (γ). Patents that are more highly cited are thought to provide the basis
for future technical advance and often change the bases of competition between firms and
industries. Hence, we might consider highly cited patents as technical breakthroughs.
Besides measuring the impact of individual patents, citations have also been used to
measure status and deference [47], knowledge flow [48], and to disambiguate inventor
careers [35]. They have also been shown to correlate weakly with financial value as
reported by a survey of patent holders [49].
While very popular, we recognize that patent citations are an imprecise and flawed
measure of technological importance and breakthrough [50]. Selecting a like patent set
for comparison and evaluation of anomalies is essential because it has been observed that
different industries have different citation patterns and norms [42] (e.g., the
pharmaceutical industry is a highly collaborative and citing industry). Also, since
citations accrue to a patent over a period of several years after it issues, citation counts
for recent patents are disadvantaged compared to older patents, an effect known
as “truncation” [42], and this is true with energy patents as well (Figure 1). In this figure,
traditional energy (10,442 patents) and renewable energy (10,603 patents) patent sets
were identified by keyword search; the random baseline includes 10,000 randomly
selected patents from the entire U.S. patent dataset.
6