This research uses social network analysis to develop models of regional innovation clusters using data from patent applications and other sources. These new models are more detailed than current industry cluster models, and they reveal actual and potential relationships among firms that industry cluster models cannot. The network models can identify specific clusters of firms with high potential for manufacturing job growth where business retention and expansion efforts may be targeted. They can also identify dense clusters of talent where innovation and entrepreneurial efforts may be targeted. Finally, this research measures relationships between network structure at the time of patent application and manufacturing job growth in subsequent years. This will permit the translation of a wide range of network-building activities into the ubiquitous “jobs created” metric. These new tools will help economic developers focus resources on high-yield activities, and measure the results of networking activities more effectively.

1. Network Models of Regional
Innovation Clusters and their
Influence on Economic Growth
Evolving the Regional Innovation Cluster Paradigm
for an Innovation Driven Economy
C. Scott Dempwolf, PhD
Research Assistant Professor
& Director
U.S. Economic Development Administration
September 4, 2014
UMD – Morgan State
Center for Economic Development

2. Regional Innovation Clusters (RIC)
“A geographically-bounded, active network of similar, synergistic or complementary
organizations in a sector or industry that leverages the region’s unique competitive
strengths to create jobs and broaden prosperity.” (EDA, 2011)
This research…
• Validates current cluster
theory and policies
• Exposes their limitation
• Offers useful extensions
We have an opportunity to take
Regional Innovation Clusters to the
next level – to respond to
innovation-driven growth
Limits of Cluster Analysis
1. Active networks are not geographically
bounded
2. presence of networks is assumed but
not measured
3. NAICS-based clusters not sensitive to
emergence
4. based on employment data; does not
connect economic growth to
innovation
5. Backward-looking

3. Innovation Driven Growth
a simple stylized model
Basic Research
Development
Invention
Product Improvement
Production
Research Parks
Incubators
Production Employment
Business Attraction
Business Expansion
Business Retention
Innovation
based
Production
TBED
Response
years

4. Innovation Driven Growth
how do we measure it now?
Basic Research
Development
Invention
Product Improvement
Production
Research Parks
Incubators
Production Employment
First Employment Data Available
Clusters defined by established
Industries, not emerging technologies
Business Attraction
Business Expansion
Business Retention
~ 5 years +/-
Innovation
based
Production
Regional
Cluster
Analysis
years
Bottom Line
Industry clusters -by whatever name- reflect
the state of innovation about five years ago

5. Innovation Driven Growth
gaining early actionable intelligence
Basic Research
Development
Invention
Product Improvement
Production
Production Employment
First Employment Data Available
This approach can shorten
the lag between real-time
innovation and actionable
economic development
intelligence by several years
while also revealing rich
talent pools, emerging
technology trends, and
specific E.D. targets
Clusters defined by established
Industries, not emerging technologies
Research Parks
Patents
SBIR Awards
NIH Awards
NSF Awards
NASA Awards
~ 5 years +/-
Innovation
based
Production
A New
Approach
Available Data Sets
Innovation
Network Analysis
years
~ 4 years +/-
State Investment Data
Incubators
Business Attraction
Business Expansion
Business Retention

6. Networks & Network Models
Georgia Innovation Network 2008 – 2010
Locations of selected actors
Networks made up of nodes
(vertices) and links (ties, edges)
Nodes are actors, agents or objects
People, Organizations, Agencies,
Documents, Places *
Links are the relationships that
connect the nodes
Regional innovation clusters are
geographically concentrated but also
have important ties to distant actors

7. Analyzing Regional Innovation Networks
Extract relationships from patent and
research grant data - about 7M records
Use social network analysis (SNA) to
analyze and visualize network structure
Theoretical grounding in sociology
and science of complexity
Behavior of the core network guides
behavior of whole network
Clustering based on intensity of
relationships
This reveals emerging technologies - what
people and firms are working on – and
specialized talent pools
Battelle Innovation Network 2005 – 2010
Created with NodeXL

8. 1. Innovation is more global and more
interconnected than previously
thought
2. Network structure influences
manufacturing employment growth
within about 3 years of patent
application (more for med & pharma)
3. Economic development strategies
that enhance innovation networks
may be a cost-effective alternative to
current capital intensive strategies.
4. Innovation networks are (or could be)
drivers of economic development in
tier 2 manufacturing regions.
PA Innovation Clusters
Westinghouse
Westinghouse cluster, Pittsburgh PA
Dissertation Conclusions
Network graphics created with NodeXL
Allegheny County
Westmoreland County
Core
2nd tier
3rd tier

9. Regional Innovation Clusters
are Complex, Emergent Systems
Networks are Ideal for Modeling Complex Systems:
• involve many interconnected or interacting parts
• exhibit emergence - behaviors that cannot be understood or
predicted by looking at the components of the system alone
• Emergence is characteristic
of self-organizing networks
• The behavior of the whole
network is driven by the
behavior of the core
• Thus we can focus on the
core and filter out the noise
Pennsylvania Innovation Networks 1990 - 2007 in the periphery

10. Applications
1. Illinois Battery Cluster (2014)
 Identifying emerging opportunities
 Combining cluster and network analysis to develop targeted strategies
2. Great Lakes Patent network (2011)
 Finding current opportunity for growth in large active clusters
 Identifying talent pools
3. Georgia Tech Research Network (2013)
 Visualizing & managing the research portfolio
 Identifying University collaborations
4. Maryland Innovation Network (2011)
 Biotech & Pharma – differentiating comingled clusters
 Zooming in to look at Baltimore’s Clusters
5. Startups, Venture Capital & Accelerators (2014)
 The CrunchBase network for Maryland
 The CrunchBase network for Illinois
6. New Jersey Solar –PV Research & Manufacturing network (2012)
 Visualizing the university – industry gap
 Developing a targeted strategy

11. Illinois Battery Cluster
The Illinois Battery Cluster
illustrates how network analysis
can augment industry cluster
analysis by identifying emerging
technologies and opportunities
for innovation – led growth.
Using network and cluster
analysis together economic
developers can rapidly develop
detailed strategies, identifying
the specific firms, institutions
and agencies involved and how
they need to connect to achieve
economic growth.

12. Illinois Battery Cluster
Cluster Analysis
• Battery manufacturing split
between two clusters
• Communications (335912: 1
est; empl N/A for 2012)
• Lighting (335911: 11 est; empl
N/A for 2012)
• NAICS 33591 Battery manufacturing
2013 (BLS)
– 2013 Location Quotient = .34
– 18 establishments
– 565 employees
Communications Cluster
Communications Equipment Components sub-cluster
Lighting Cluster
Storage Batteries sub-cluster
Conclusion: Limited opportunity

13. Illinois Battery Cluster
Network Analysis
• 2012 - $120M JCESR created at Argonne
• ‘5-5-5’ goal  significant industry growth
~ 2017
• Network identifies specific firms + real &
potential research ties in specific
technologies
Conclusion: The combination of limited
production capacity (from cluster analysis),
strong research capacity & research investment
suggest specific economic development
strategies to capture future job growth.
• Build industry partnerships around existing
firms & supply chains to facilitate growth
• Target specific firms for attraction to grow
cluster rapidly

14. Great Lakes Regional Innovation &
Manufacturing Clusters (core)

15. Great Lakes Innovation Clusters Impact
on Planning Practice

16. Potential University Applications
Well suited for integrating and
managing research across
multiple institutions via open
networks rather than institutional
structure
Visualizing a Research Portfolio
Offers both a big picture and
details of technology
commercialization areas and
opportunities Georgia Tech Innovation Network 2008 - 2010 (2 steps)
Created with NodeXL

17. Maryland Innovation Clusters 2008 - 2010
This analysis showed that the
clustering algorithm is sensitive
enough to distinguish between
pharma and biotech.
Maryland Innovation Network 2008 – 2010
Created with NodeXL
Baltimore Innovation Network 2008 – 2010
Created with NodeXL

18. CrunchBase startup networks 2005-2014
Illinois Startup Network
Although similar in size the
Illinois network exhibits more
robust structure
Little discernable structure; clustering
appears weak
Maryland Startup Network
Some structure and beginnings of
clusters apparent

19. Maryland
Startup Network (CrunchBase 2005 - 2014)
When clustered, spatial agglomeration is the main organizing factor
both locally and for distant capital sources; DBED & TEDCO feature
prominently, followed by a few investment firms.

20. Maryland
Startup Network (CrunchBase 2005 - 2014)
Removing New York, Boston and San Francisco nodes diminishes
spatial influence as an organizing influence, allowing technology
clusters to emerge. (between-cluster ties hidden in this graph)

21. Illinois
Startup Network (CrunchBase 2005 - 2014)
Spatial agglomeration is an important factor in Chicago and North
Shore clusters; Excelerate Labs & HealthBox are prominent accelerators
locally; Strong ‘portfolio’ organization in remaining clusters

22. New Jersey Solar PV Cluster 2008 - 2010
Fruchterman-Reingold layout
In NodeXL
This analysis revealed significant
gaps between solar PV research &
development and solar PV
component manufacturing.
Grid layout in NodeXL
Production Core
Research Core

23. Next Steps
Academic Research
• Publications
• Presentations at SSTI, TCI Global
• Complete County-level application – St. Mary’s
County, MD CEDS
• Seek NSF SciSIP funding for additional
network research; validation & calibration of
the economic model
• Pending proposal with NIST to evaluate their
impact on innovation and commercialization
(alternative metrics to patent counts)
• Collaboration with UMD HCIL on
improvements to visualizations and NodeXL
software
Commercialization
• Launch startup company (fall 2014)
• Engage ten pilot communities / regions over
the next two years
– Mix of different sizes, scales, level of organization,
density
– Focus primarily on manufacturing regions
– Some with cluster strategies, some without
• Pilot studies may include
– A network report (limited version of Illinois
Roadmap)
– Traditional cluster analysis using the Harvard
tool for regions that don’t have it
– An interactive network model
– On-site training & Technical Assistance
• Evaluation of performance across all pilot
regions