Page 12 -

this first exploratory work. Follow-on work will provide more comprehensive
investigation of the historic social variables involved in energy innovation.

2.2 Patent Interrogation
Three independent variables were developed to provide insight into the structural origins
of inventive activity: “Government Interest,” “Assignee Organization Type,” and
“References to Science.” Government interest patents are those that the government
retains a right to, usually because they funded the work. Patents with government
interests were identified using a USPTO search tool [35] to search the government
interest field for identifying terms such as “defense” or “energy.” The search terms could
not be as rigid as “Department of Defense” because often the patents cited the “Dept. of
Defense,” so search terms were developed to be inclusionary. Assignee organization type
defines what kind of organization patented the invention. Text search algorithms were
developed to search the patent assignee field for terms (e.g., “university,” “corporation,”
and “LLC”) that identified the organizational type of the inventing institution. Assignees
were classified into five types: Private, University, Government, Independent (i.e., no
assignee), and Other. The results of the automated text classification system were then
manually reviewed for accuracy. References to science were identified if at least one
other non-patent reference in the “Other References” field was coded as referencing
science (these references are mostly to scientific publications [40]). While the variable
for references to science was included as a covariate control in our model, its influence
was not simulated for this investigation.

Three independent variables were developed to provide insight into the social origins of
inventive activity: “Team Size,” “Mentorship,” and “Dispersion.” Team size counted the
number of inventors listed on the patent application. Patents with two or more inventors
were classified as team patents. This variable was included in our simulation results.
Mentorship compared the time-in-field of inventors on a patent. If a difference of 10
years or greater existed between any two inventors, the patent was classified as arising
from a mentorship relationship. Time-in-field was derived by subtracting the application
year of each inventor’s first patent from the application year of the patent in question.
Dispersion measured the maximum distance between any two inventors on a single
patent, with a distance greater than 100 miles classified as dispersed. Distance between
inventors was calculated using the zip code values corresponding to the Inventor City
field and so returned values only for patents with two or more U.S. inventors. Mentorship
and dispersion variables were included as covariate controls but were not simulated for
this manuscript. We also performed qualitative analysis of U.S. inventor patent co-
authorship with collaborators in other countries for work unrelated to this manuscript and
demonstrated the ability to map collaboration between the United States and other
countries. Further research should include work to understand the impact of non-U.S.
innovations on domestic innovation, both from the perspective of contribution to the
collective knowledge as well as more direct individual collaboration effects.

2.3 Patent Value Definition
2.3.1 Technical Importance: Citation Analytics
Citations have a long history in the bibliometric literature and the first application to
patents identified was by Trajtenberg [34]. Heretofore, patent citations are the dominant

5