Page 74 - Technology and Innovation Journal - 19-1

412 DEMPWOLF & SHNEIDERMAN

• Do regions with higher innovation network medical devices because they are regulated and
density innovate faster? What network struc- tested by product name. Otherwise, prod-
tures are associated with faster innovation? ucts are typically not identified in STI data
sources. One data source that associates product
Both are active research questions for the authors. names with the firms that produce them is the
Regarding accelerators, a 2014 study of innovation UPC database. The dates associated with UPC
accelerators for the U.S. Small Business Adminis- records are the date the record was last updated
tration found no good metrics in the literature that rather than the date of product launch, but the
answered the question of whether accelerators did source is worth further investigation.
indeed accelerate innovation (7). A subsequent 2. STI data resides in multiple unlinked admin-
network analysis comparing outcomes between 77 istrative databases, and data quality is vari-
accelerator-affiliated start-ups and 77 non-accelera- able. Data cleaning, matching, and disambigu-
tor-affiliated start-ups receiving angel funding found ation is a significant, time-consuming, and
that the accelerator subnetwork was 8.5 times larger ongoing task. Records are not always complete,
than the unaffiliated angel network and exhibited and augmentation may be necessary. Efforts
more opportunity for brokerage. Accelerators invested to automate data preparation processes through
33% less per start-up in angel funding ($100,000 vs. machine learning and other algorithms are
$150,000) and 50% less overall ($1.3 billion vs. $2.6 underway, but this will still take time.
billion) than unaffiliated angels. Combined, their 3. Innovation processes comprise many differ-
start-ups raised an additional $41 billion in subse- ent events, and those events may involve dif-
quent funding rounds and acquisitions (7). While ferent networks of people and organizations.
these results suggest that accelerator-affiliated start- Finding the relationships among events is not
ups may be more efficient, they do not answer the always easy.
question of whether the accelerator-related start-ups 4. Technology topics have not been standardized
achieved those results faster than non-accelerator across the various types of events although there
start-ups. A pending EventFlow offers the poten- have been numerous advances in topical anal-
tial to answer that question using the same dataset ysis and natural language processing.
(CrunchBase) as the 2014 study. 5. Data remains incomplete.
The question of whether regions with higher net- 6. FDA drug databases and medical device data-
work density innovate faster was recently embedded bases are structured differently and contain
in a successful funding application for the National different information. For example, medical
Institute for Innovation in Manufacturing Biophar- devices may be linked to clinical trials, but
maceuticals (NIIMBL) under the National Institute there are no linkages between drugs and clinical
of Standards and Technology. The authors will use trials. Drugs may be linked to patents, but there
EventFlow and NodeXL to model the network struc- are no linkages between medical devices and
ture and innovation outcomes of NIIMBL partners patents.
and others in multiple regions throughout the U.S. 7. Applying this methodology to other critical
over the next five years to answer this and other industry sectors may be useful. Clean technol-
related questions.
ogy and energy, for example, share many sim-
ilarities with medical devices in terms of
CURRENT DATA LIMITATIONS inputs, outputs, innovation trajectories, regu-
As promising as the preliminary results are, several lations, and challenges. The Lab-to-Market
data limitations are hindering broader application of initiative and the Department of Energy’s Office
this temporal analysis technology to understanding of Energy Efficiency and Renewable Energy
and measuring innovation processes: may offer comparable data to help overcome

1. Data is typically not collected or organized the identified data challenges.
around products as the end result of innova-
tion. Product data is available for drugs and