Studying Network Effects In Complex International Science, Technology and Innovation Partnerships: A Case Study Of The MIT Portugal Program Mac Hird Engineering Systems Division, MIT Committee: Dava Newman Cesar Hidalgo Sebastian Pfotenhauer June 24, 2014 1 Complex International STI Partnerships: A booming phenomenon • Governments are increasingly utilizing international partnerships to build STI capacity • Focus: Universities • Student exchanges, joint research projects, dual degrees, or branch campuses… • New complex forms of partnerships emerging, that go beyond previous efforts: • Portugal: International Partnership Program • MIT, CMU, UT Austin, Harvard, Fraunhofer • Singapore: CREATE Campus, SUTD • MIT, Cambridge, ETH, TU Munich,… • Middle East • Masdar, KAUST… 2 Government Rationales for International University Partnerships • Transition to innovation and knowledge-based economies • Universities are key: Simultaneously address… • Human research development • Research • Technology development, innovation, entrepreneurship, tech transfer • Institution-building, cultural change… • International linkages are key: • Integration into knowledge networks (“globalizing learning economy”) • Research networks increase productivity • Global competition: adoption/adaptation of international best-practices • International visibility & branding: attract best and brightest • Support Institutional and Cultural Change 3 Why Study these Collaborations? • New policy instrument • Pool expertise from external organization to build domestic capacity • They are a new type of “tech transfer” • • Transferring organizational and scientific practices rather than physical technology Large investments of capital • Understudied: primarily practitioner-driven • Broader lessons for collaborative/open innovation and economic development 4 Pilot Case study: MIT Portugal Program (MPP) • • Multi-pronged international partnership between 6 PT universities and MIT • Collaborative research in four focus areas • 7 inter-institutional graduate programs • Collaborative innovation and entrepreneurship activities • Phase 1 (2006-12), Phase 2 (2013-17) 4 main objectives (among others) : • Encourage PT universities to work closer together (MIT as incentive/”glue”) to build critical mass • Encourage Portuguese collaboration with MIT • High-impact research • Increase PT visibility and attractiveness 5 MPP Systems Architecture Portugal From Pfotenhauer, Roos and Newman, 2013 6 MPP Systems Architecture Portugal MPP MIT From Pfotenhauer, Roos and Newman, 2013 7 MPP Systems Architecture Portugal MPP MIT From Pfotenhauer, Roos and Newman, 2013 8 Research Question • How do the collaboration networks of researchers change through the adoption of the MIT Portugal Program? • • Has the collaboration network developed as policymakers have expected? Future Questions: • How does the impact of CISTIPs compare across fields, institutions, and countries? • How do collaboration networks and network dynamics reflect partnership architectures? • How does this fit into national policy trajectories? 9 Data Sources • Researcher-centric Collaboration Network • 297 MPP-Participating Faculty • Articles, Conference Proceedings, Books, and Book Chapters in • • • ISI Web of Science Scopus (by Elsevier) Focus on 1996 – 2014 10 Control Group • 100 Non-MPP Portuguese Researchers • 4 universities (IST, Porto, Minho and Coimbra) • Select non-MPP faculty in Chemical Engineering and Mechanical Engineering • Similar in age and pre-2006 number of publications • Not involved in the CMU or UT Austin collaborations in Portugal 11 Confirming expectations: Portugal on the rise Nearly 20% increase in the number of publications per faculty above the control group for those participating in MPP • Yearly Number of Publications Normalized Yearly Number of Publications MPP Average Publications per Researcher/ Control Group Publications per Researcher Difference between MPP and Control (Publications per faculty per year) 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 1995 2000 2005 2010 MPP Begins 2015 1.4 1.2 1 0.8 0.6 0.4 0.2 0 1995 2000 2005 2010 2015 MPP Begins 12 Effect on New and Experienced Researchers Much larger effect on New Researchers, who are more than twice as productive on a yearly basis as their control group counterparts • Normalized Yearly Number of Publications 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 1995 -0.2 2000 2005 2010 -0.4 -0.6 MPP Begins 2015 MPP Average Publications per Researcher/ Control Group Average Publications per Researcher Difference between MPP and Control (Publications per Researcher per Year) Effect Size on New and Experienced Researchers 3 New Researchers (<4 Years Publishing in 2006) 2.5 2 1.5 Experienced Researchers (>15 Years Publishing in 2006) 1 0.5 0 1995 2000 2005 2010 2015 MPP Begins 13 Effect on Existing Collaborations • What happens to a researchers existing network? • • Do new connections caused by MPP augment a researcher’s collaboration network or replace it? MPP Collaborations replace some existing network connections Collaboration Persistence MPP Contro l Percent of collaborations in 19952000 that will also collaborate in 20012006 26.9 21.5 Percent of collaborations in 2001-2006 that will also collaborate in 2007-2012 35.5 47.3 2001-2006 2007-2014 14 Effect on Intra-Portugal Networking • While there has been an increasing trend in Portuguese-Portuguese connections, MPP has accelerated this trend New PT-PT Connections in MPP/ New PT-PT Connections in Control Normalized New PortugalPortugal Connections per Person Portugal MPP 1.8 1.6 1.4 1.2 MIT 1 0.8 0.6 0.4 0.2 0 1995 2000 2005 2010 2015 Year MPP Begins 15 Intra-Portugal Networking 2001 - 2006 MPP Portugal MIT 16 Intra-Portugal Networking 2007 - 2012 MPP Portugal MIT 17 Spillover to broader collaboration network MPP faculty are not only collaborating with other MPP faculty, but also with the networks of other MPP faculty • Difference between MPP and Control (% of new connections per year) Percent of New Connections through Introduction No Introduction Introduction MPP MPP Non-MPP 14 Non-MPP 12 10 8 6 4 2 0 1995 -2 -4 2000 2005 2010 2015 MPP MPP Year MPP Begins 18 Conclusions • The MIT Portugal Program: • Led to more publications with a slightly higher impact factor • Led to more collaborations, both within Portugal, with American researchers and with researchers around the world • Has developed structurally as policymakers have expected • Develops connections which replace some existing relationships • Has positive spillovers to non-MPP faculty 19 Future Work • Statistical Analysis of Network • Include other types of MPP interactions • Informal collaborations, acquaintance networks • Qualitative Data • • To understand why and how particular relationships are formed Match with systems architecture analysis • Other international partnerships • • Including the other MIT collaborations and others from around the world Extend research to non-university partnerships 20 Looking at Different Partnerships Architectures From Pfotenhauer, Roos and Newman, 2013 21 Thank you 22 Control and MPP Degree Distribution MPP Control 23 Control and MPP clustering coefficient MPP Control 24 Control and MPP Neighbourhood Connectivity MPP Control 25 Control and MPP 26 Control and MPP 27 Control and MPP Network Measures MPP Control Clustering Coefficient 0.831 0.821 Connected Components 51 20 Network Diameter 10 9 Network Radius 1 1 Network Centralization 0.089 0.066 Characteristic Path Length 4.226 4.175 Avg. Number of Neighbours 11.160 10.116 Number of Nodes 21815 7710 Network Density 0.001 0.001 Network Heterogeneity 2.187 1.767 Analysis Time (sec) 28035 1139 28 MIT International Collaborations 29 New International Collaborations • Large initial increase in collaboration between USA and Portugal due to MPP, but over time the diversity of collaborative connections has increased • “New International Connection” – First time collaboration between two researchers, even if they are connected elsewhere in the network Percentatage of New International Collaborations 100% 1600 90% 1400 80% 1200 70% Percentage 1800 1000 800 600 All Others Italy 60% Brazil 50% France 40% Spain 30% 400 Germany 20% 200 United Kingdom 10% 0 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 New Collaborations per Year New International Collaborations by Year Year MPP Begins 0% 199619982000200220042006200820102012 United States Year MPP Begins 30 New International Collaborations New International Collaborations by Year 1800 New Collaborations per Year 1600 1400 United States 1200 United Kingdom 1000 Germany Spain 800 France 600 Brazil 400 Italy All Others 200 0 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Year MPP Begins 31 Effect on Intra-Portugal Networking Percentatage of New International Collaborations 100% 90% 80% Percentage 70% All Others Italy 60% Brazil 50% France 40% Spain Germany 30% United Kingdom 20% United States 10% 0% 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Year MPP Begins 32