Page 35 -

These empirically derived findings contribute to the dialogue on where and how
innovations, and energy innovations in particular, originate and how we can create more
of them. In particular, it has been widely accepted that collocated communities of
practice—universities, corporations, and venture capital—are helpful and perhaps
essential for creating vibrant innovation ecosystems [82, 83]. Indeed, in many industries,
such as those related to energy including biotechnology (biofuels) and semiconductors
(solar), the areas of Silicon Valley [84, 85] and Greater Boston [82, 86] are indicated as
innovation and company hotbeds. However, these data demonstrate that indentified
biofuels patents from California are mildly discounted when compared to all other
patents. Similarly, the patents from Massachusetts, while perhaps technically important,
have limited commercial importance. We realize the limitations of these interpretations
and hesitate to extend these observations much further without significantly more
investigation. However, unique technology development and deployment dynamics exist
within the energy sector due to the nature of the technologies and markets. These unique
differences require us to reevaluate how public and private research, development, and
deployment capital is used.

Biofuels is our smallest dataset, with only 938 patents. Each of the geographical, social,
and institutional patent sub-populations is proportionally limited, so finding statistically
significant correlations between the patent populations is more difficult than in the solar
and wind sets. Furthermore, the work of prescribing reliable funding strategies becomes
more tenuous as the sample size decreases, so we will focus our preliminary prescriptive
efforts on the richest dataset available—solar patents. However, we expect to be able to
add prescriptive detail as patent sample sizes increase through improved search methods.

3.4 Predicting Breakthroughs
Through a retrospective analysis of energy patents, we have demonstrated the ability to
ascertain key drivers of breakthrough technical and commercial impact. The introduction
of the new measure of screened Web hits as a predictor of commercial value shows
significant promise as a tool for bibliometric researchers. The logical next step is to
determine if this type of analysis can be used predicatively to anticipate breakthrough or
commercially impactful innovations. While we are admittedly a long way from being
able to do this with any assurance of accuracy, we define the following equation in an
attempt to begin a thread of inquiry wherein scholars and practitioners alike can
anticipate the value of a patent. For the licensed and unlicensed NREL patents, we
calculate the following ratio for each one:

κ = (ω sig − ω sig ) (ω+ agg − ω agg )

where κ is the calculated potential impact of the patent, and the average signal or
aggregator Web hit value is subtracted from the unique patent Web hit value. If κ for a
patent is positive, that patent is more likely to be impactful or commercially relevant.
Indeed, for our two populations of patents, 62% of the licensed patents are correctly
tagged positive while only 30% of the unlicensed patents are. Furthermore, comparing
the average commercial potential for the licensed κ ( ) versus unlicensed κ ( unlic ) patents
lic
by year shows that the relationship predominantly correctly predicts commercial value

28