Rick Cleary, Brigitte Muehlmann, and Davit Khachatryan, Babson College

Many statisticians with active research careers learn to demonstrate that their scholarly work has been of interest to others by keeping track of variables such as the number of papers written, number of co-authors, and number of citations in the literature. Annual reports to administrators often include more sophisticated measures about journal quality, including impact factor, rejection rate, and circulation.

Internet tools like Google Scholar and ResearchGate have made the work of collecting this information easier than ever. Statisticians and researchers know these numbers are imperfect and sometimes misleading, but they are expected to produce them because they have value in providing department and university leaders with data to show funding agencies, trustees, and legislators that they are producing something worthwhile.

There is an often overlooked way in which research of all kinds, and statistical scholarship in particular, contributes to the advancement of science. For example, what do the following three papers—published in respected outlets—have in common?

• “Shrinkage Estimation of Price and Promotional Elasticities: Seemingly Unrelated Equations” by Robert Blattberg and Edward George, Journal of the American Statistical Association, 1991
• “Inference in Model-Based Cluster Analysis” by Halima Bensmail, Gilles Celeux, Adrian Raftery, and Christian Robert, Statistics and Computing, 1997
• “Fisher in 1921” by Stephen Stigler, Statistical Science, 2005

All three papers have been cited multiple times in patents granted by the United States Patent and Trademark Office (USPTO); in fact, the paper by Blattberg and George has been cited 27 times.

There are dozens of articles published in ASA journals that have been cited in patents. The work of the statistics community is informing innovation in numerous ways many are unaware of.

The Patent Process

How does the patent process work in the United States? Why would someone cite a statistics article, especially an expository article, in a technical report about an invention?

The patent process has many parallels with the publication process academics know well. An invention is submitted and then patent examiners, loosely equivalent to referees and editors from the academic world, decide the fate of the patent application. The application might be issued but, if not, the patent examiners provide feedback and the applicant may choose to revise and resubmit. Like journal submissions, the time to examine patent applications can vary substantially. According to the 2019 USPTO Performance and Accountability Report, the average first action pendency (i.e. equivalent to when a journal editor notifies the author of the results of the first round of review) is 14.7 months, while the average total pendency is 23.8 months.

Patents are granted to inventions that are novel, non-obvious, and useful, assuming the invention is described adequately in an application. To establish novelty, all the relevantly available public knowledge—called “prior art”—needs to be acknowledged through citations. Pre-existing patents and published patent applications, as well as non-patent literature (e.g., journal articles), are expected to be cited by the applicant in a patent document to establish the novelty of the application. Citations can be, and often are, required by the patent examiners as part of their decision.

More so than additional citations in the academic publication process, an examiner’s addition of prior art citations has consequences for inventors. It reduces the scope of patentable inventions.

It is important to note that being cited in a patent does not necessarily mean something the author wrote has contributed to a major commercial success. Most granted patents go on to live quietly with little impact on the particular business or industry they are associated with. As with journal articles, it can be difficult to predict which individual efforts will come to be widely recognized as important. So, while it is not possible to put a dollar figure on the value of statisticians’ work, what the many patent citations of statistical articles do show is that statisticians’ academic work adds value to building a useful invention.

Has Your Work Appeared on a Patent Document?

The USPTO provides an open access search tool that helps identify potential patent citations to academic work. Non-patent literature is searched using the field name pertaining to “Other references,” which is coded as “OREF” in the search tool. To find out if there are patents citing an article, use the “OREF” keyword with the first author information and one or more representative terms from the title of the article.

For instance, to arrive at the citations for the Blattberg and George (1991) article on shrinkage estimation, the following search string was used: OREF/(blattberg and shrinkage). To increase the precision of retrieved results, one may have to fine-tune the search string by providing additional representative terms from the title of the article.

When it comes to author names, it may be important that only the first author’s information is provided (e.g., Blattberg). This is because USPTO does not use a consistent citation style. Patents may cite an academic work only using the first author’s information followed by “et al.” and journal information.

Implications

Professionals who have published work cited in patents have a tangible way to demonstrate the utility of their ideas. The fact that some of the statistics articles cited in patents are in expository journals is an argument in favor of good writing. Patent examiners must learn the basics of many fields, so it is not surprising they might find a well-written overview of a topic to be more useful than a highly technical paper in the same field.

There are many ways that discovering a patent citation could be useful for individual statisticians—and for the profession more broadly. Among the most important are the following:

• Encouraging students: It’s hard to imagine a better answer to the ubiquitous student question, “What’s this stuff good for?”
• Performance reviews: An academic statistician whose work is cited in a patent has a convincing argument to show a chair or dean that they are making a useful contribution. Chairs and deans, in turn, have the raw material to persuade trustees or legislators that funding statisticians’ work has helped inform the creation of useful inventions worthy of a patent. These groups that hold academic purse strings might especially value patent citations as they may find the quality of academic publications difficult to ascertain.
• Promotion of the ASA and statistics: If statistics journals are fueling innovative work in various parts of the economy, statisticians have much greater credibility when they approach granting agencies for financial or program support.