Key lessons from the GIC 2019
… And what’s next
Thanks to the collaboration between YSI INET, Utrecht University, Institutions for Open Societies (IOS), Centre for Complex Systems Studies (CCSS), the European Commission, TREND and the Collective Learning Group, MIT Media Lab, we had another great conference this year.
What an experience, both in terms of intellectual stimulation and learning and in terms of social affinity. There was a refreshingly high concentration of innovative ideas and conclusive discussions. It can be established without a doubt that this conference contributed to the formation and connectedness of a network of scholars that is apt for advancing the science in the field of economic geography, innovation and complexity.
13 keynote speakers, 31 young scholar presenters and 12 poster presenters from 5 continents took us into the most ingenious approaches to the hottest topics on the research frontier. We were guided from complexity as a conceptual paradigm to complexity as a network property to complexity as an operational science. Apart from complexity and innovation, we were presented new insights into relatedness, resilience, inequality, sustainability, the new economy, knowledge, the labor market, economic policy and cities. Thereby, the manifold formats, from presentations to panels, from poster sessions to the final machine learning workshop and the occasional coffee breaks allowed for different modes of engagement.
Throughout the conference, some of the greatest professors in our field of science made the case for paradigm shift, presented groundbreaking findings and pointed towards future research avenues. For you to play around a little more with these ideas, I have summarized a few key takeaways below.
See here fore the API tutorial and the Web Scraping tutorial presented by Alex Bishop and Konstantinos Stathoulopoulos from Nesta.
Click here for the introductory post.
Funding emerging Ecosystems
In the first lecture after the big picture discussion, Maryann Feldman pointed towards the contingencies of firm’s response to capital. Tracing the emergence of the North Carolina Research triangle, she finds that there are three stages of development defined by the inflection points of firm density as a function of time. In her study she evaluates the individual and joined impact of three sources of finance on firm survival in each stage of development — private venture capital, smaller state funds and quality-based federal government funds.
Her findings show that depending on the context, particularly the stage of development and type and mix of capital, firms respond differently. In general federal and private funding significantly increase the firm survival rate. However, there is more nuance to it. State and private capital pay off better in later stages of development and more sources of funding decrease the probability of a firm dying.
See here for the paper reference.
Lastly, she repeatedly stressed the importance of acknowledging entrepreneurial agency.
“A place that is willing to build success is able to do so”
The Supply Chain Economy: Understanding Innovation and Growth in Services
Mercedes Delgado started her speech by pointing towards a paradox. Despite dramatic decline in employment, innovation economists have an obsession with the manufacturing sector. To adjust to the new economy, she proposed a new innovation framework, whereby the focus lies on the suppliers of inputs and trade services instead of consumer products — the supply chain economy.
The latter is important for innovation in three ways: (1) it produces specialized input that is integrated into the value chains of business; (2) it has more downstream linkages for diffusion and recombination; and (3) can capitalize on agglomeration benefits. Moreover, the supply chain economy is characterized by a high level of innovation, decent wages, the largest percentage of STEM jobs and strong economic growth. Hereby, she importantly noted that patents are biased indicators for innovation, because services are difficult to patent.
Despite its potential, the supply chain economy bears its own dangers. Mercedes Delgado is especially worried about its contribution to job polarization, but also about its vulnerability to crisis. For the latter, at least clustering can increase resilience as her paper shows.
See here for a summary of her Chapter in the New Oxford Handbook of Economic Geography.
Complexity and Policy: Searching for relevance
During her agenda setting talk, Elvira Uyarra has posed a set of key questions for us regarding the role of policy and its complexity for industrial diversification.
What role do institutional settings have and when do they matter the most?
Is there such a thing as policy relatedness and how would we conceptualize it?
How can we integrate the demand side into smart specialization to tackle societal challenges?
How do policy worlds evolve over time and in what way are they conductive to the emergence of a phenomenon?
How can we give recommendations while taking into account the political economy?
Lastly, she stated that policy complexity is mainly defined by agency, the nature of instruments, their high degree of interpretive flexibility, the multiplicity of policy interactions, time, the impact of the political, cultural and institutional context and its implementation capacity.
See here for her paper on the role of public procurement for innovation policy.
Regional Diversification
Ron Boschma reminded us that local capabilities are opportunities but also limits, which is why places are likely to diversify into activities that are related in terms of skills, value chains or inputs. Unrelated diversification on the other hand is rather the exception, but can pay off immensely when leading to the introduction of more complex activities that are harder to imitate.
As related diversification can lead to lock ins, strong and experimental policy interventions are sometimes necessary for enabling the creation of new capabilities as basis for the lucrative unrelated diversification. This risk- benefit trade-off is the core challenge of the Smart Specialization Strategy that the EU adopted as part of its innovation policy. Hereby, Boschma stressed that the regional diversification strategies need to be adjusted to the local context and capabilities.
In his speech he posed a new question for regional science:
“How far can you import capabilities from elsewhere that you miss in your economy for new industry exploration?”
The possibility to absorb external knowledge through inter-regional linkages to avoid regional lock ins and venture into unrelated diversification has not been fully exploited yet. To aid this new approach, he proposed the mapping of technology specific complementarities. He and some of his colleges will publish a paper soon that shows the significance and terms of this new spatial framework for regional diversification.
Percolation as a diffusion model of new products and new ideas
In an extraordinary Tuesday morning lecture, Koen Frenken demonstrated that social network structure matters for innovation diffusion and presented his model for diffusion efficacy.
“Standard percolation models translate directly into models of product demand in economics and diffusion of ideas in science studies”
In his model the reservation price, the individual max price at which a person wants to buy the product, is distributed heterogeneously throughout the social network. At each price of product the social network splits into people willing to buy the innovation or not. The percolation threshold and maximum efficiency is defined by the highest price which still allows for diffusion.
In the idea diffusion model, ideas get adopted when they meet the persons minimum quality requirement. Incorporating the mechanism of social reinforcement, Koen Frenken adjusted the model so that higher numbers of neighbors in the social network adopting the idea lowers the person’s minimum quality requirement.
Running the models with different network structures, he finds that clustering, a small world characteristic, can inhibit complex propagation by locking in information distribution. In fact, small world networks never outperform random networks, even if social reinforcement is present. Moreover, the more preferences are homogeneous the less the structure of the network matters.
The long term trajectory of the Western European economy 1000–2000 in global perspective, with special focus on the role played by institutions
With a fresh outside perspective, the economic historian Jan Luiten van Zanden reminded us that Europe experienced an economic miracle when its knowledge economy took off from the 15th century onward despite China having had a much wealthier and knowledge intensive history.
He cites the example of printing, which was invented in China in the 11th century. While the number of printed books - the manifestation of knowledge at that time - in China remained relatively constant for centuries, in Europe the number of books rose rapidly and exponentially when it appropriated the technology much later in the 15th century.
For Jan Luiten van Zanden differences in the social hierarchy, institutions and power structures can to a large part explain the miracle. He set out a convincing case for why girlpower, the higher degree of female agency in Medieval and Modern Europe compared to China, contributed to the historically strong knowledge economy in Europe. Morover, Europe had an advantage through its less hierarchical more negotiation based institutional structure.
How mismatches at the labour market may obstruct the energy transition
In a very insightful presentation, Anet Weterings showed us how the labor market hampers the reduction of green house gas emissions and what is needed to reform it to prepare it for the energy transition. Her team took two scenarios of substantially greener economies as developed by colleges of the Netherlands Environmental Assessment Agency(PBL) and analyzed the gap in the labor market between the scenario and the current situation.
In particular they modeled how the demand will shift in different industries by conducting a multi-regional input — output analysis and incorporating interdependencies of industries. Therefore, they took into account the crowding out, rebound and value-chain effects that indirectly impact labor demands in different industries in case of an energy transition. However, identifying the labor demand in different industries is not sufficient as labor demand needs to be matched to jobs and jobs need to be matched by workers. Therefore, Anet Weterings and her team acknowledged that mobility across industries is limited. Based on empirical data on intersectional job flows they modeled the willingness to move to new industry.
Through their approach they were able to identify tensions between vacancies and job searchers. Although nationally the need for labor is only slightly increasing, on the regional level strong shifts across industries appear necessary. This points towards the need for reforms that are adapted to the regional context.
The role of modular structure in network-based modelling of industry agglomeration and growth patterns
In certainly the most mathematical lecture of all, Neave O’Cleary went us through two of her research projects on agglomeration network science.
For explaining why industries cluster in space and which factor is most important, she and her colleges decomposed the global network structure of co-located industries into regions of networks that are related through different types of agglomeration. Hereby, they operationalized the Marshallian externalities of labor pooling using employment data, value chain linkages using input-output tables and knowledge sharing using patent citation data. To test the stability of graph communities across time scales using the stochastic Louvain’s algorithm, they apply the Markov stability optimization criteria.
They find that labor sharing and value chain linkages are the strongest agglomeration forces. Moreover, clusters driven by labor sharing tend to me more educated, while those driven by value chain linkages are less educated.
See here for the corresponding paper.
Her second research project shows how the modular structure in labor networks of the Irish economy reveals skill basins. The Irish economy has the special feature of a long capability-wise distance between domestic and foreign firms. Therefore, the government is interested in labor transitions between industries. Neave O’Clery et al. find that while finance and computing are very isolated sectors, construction, food and manufacturing are very connected in terms of labor sharing.
Soon to come: O’Clery N, Flaherty E, Kinsella S (2019) Modular structure in labour networks reveals skill basins.
Finally, she poses a key methodological question: “How do we compare networks in a robust manner?”. She proposes that the network’s modular structure could be meaningful network property for comparison.
The geography of complex knowledge
In his speech Pierre-Alexandre Balland guided us from complexity to spatial inequality. He lays out the reality of the winner takes it all economics, which in his opinion is the result of network effects, especially preferential attachment. Through digital technologies, transport and globalization knowledge consumption becomes flatter, more and more people consume the same products/services worldwide. On the production side, increasing economic complexity, which goes hand in hand with agglomeration benefits, leads to higher spatial concentration. The super-linear relation between scaling in terms of population and patents hints at that. In fact, economic complexity may even serve as a signal that population will increase in the future.
Pierre-Alexandre Balland showed us that the world is getting more spiky, but considerably more so for very complex knowledge compared to less complex knowledge. Especially the polarization between the urban and rural world is rising. Finally, he points towards the importance of institutions as a social mechanism to counter the extreme imbalance in the system.
“How can we make sure that policies like the smart specialization strategy don't increase this great spatial divide?”
Should we and can we promote innovation in peripheries?
Venturing his yet unpublished paper, Andrés Rodrigues-Pose started by scrutinizing the conventional linear model assumption that investment in R&D would lead to more innovation, which in turn would stimulate productivity and growth. The failure of the model is epitomized by the fact that geographical polarization in Europe persists despite R&D funds specifically targeting lagging behind regions. He makes the case for why it is time to reconcile Europe’s conflicting objectives of excellence in innovation on the one hand and territorial cohesion on the other and what we need to change in our approach.
His data analysis leads to a few interesting observations that highlight the potential and special circumstances of the periphery: (1)The less developed regions of Europe, Eastern and Southern Europe, have difficulties transforming R&D in innovation. (2)Benefits of Quality of Government have the greatest impact in regions with low R&D, which suggests that these are the places where poor institutions need to be tackled. (3)While in general, patenting is disconnected from growth, there is a greater probability that patenting will turn into growth in less developed regions mainly Eastern Europe.
“The problem is one of innovation, diffusion and assimilation, not of research.”
Rodrigues-Pose claims that there is too much concentration on “research” in our R&D framework, which can be seen by the exponential increase in publications, many of which are disconnected from the local fabric. The “development” part of R&D on the other hand is neglected.
How fast is this novel technology going to be a hit?
Fabiana Visentin and her colleges traced diffusion paths of newly appearing technologies by identifying follow on technology that reuses the novel technology. Hereby two parameters of diffusion legitimation, which is the time needed to reach an initial level of diffusion, and the technological impact, when it reaches its point of saturation. In general the two desired traits of short legitimation and high technological potential have an inverse relationship.
She convinces us with that the diffusion of the technology is determined by the combined technological and cognitive characteristics of its components — their similarity, science-basis and the familiarity of the inventors. New technologies with more similar components as well as those with more familiar inventors tend to have shorter legitimation time but also smaller technological impact. More science based new technologies on the other hand need longer legitimation and have greater technological impact.
Read here for a summary of her article.
Finally, she poses the question: “What do citations really measure?”
Do Not Put Eggs in One Basket: Related Variety and Regional Economic Resilience in China in the Post-crisis Era
Taking us to the specific context of China, Canfei He convinced us that related variety can have a negative impact on regional resilience to economic shocks.
“In China inland cities were more resilient to the 2007–2008 economic crisis than regions with a high level of related variety.”
Only the supply side perspective of the effect of related variety on long term regional resilience has been tested. To address that limitation, Canfei He modeled the demand side effects of related variety and thoroughly tested their effect on short-term resilience based on Chinese export data before, during and after the economic shock. He shows robust empirical evidence that through two mechanisms related variety made China more vulnerable to external demand shocks.
Firstly, related variety has risk spreading effects through its formal and informal firm networks. Secondly, related variety has a product quality promoting effect due to knowledge spillovers, which leads to lower quantity being produced and also higher demand elasticity. Therefore, demand shocks can cause greater damage.
His findings show that the net effect of related variety is contingent and that we should better understand it as a double edged sword.
Click here to see his book on Evolutionary Economic Geography in China.
How Humans Judge Machines
In the shortest and last keynote speech of the whole conference, César A. Hidalgo brought a new, future oriented topic on the table: AI and Ethics.
With about 14 scenarios of ethical dilemmas, whereby in each scenario the protagonists was exchanged being once a human once a machine, he played around with our imagination. But instead of only provoking deep self-reflection about one’s own moral (under)standing, he guided us through the approximation of the actual method of his investigation of how humans judge machines.
Based on Haidt’s moral dimensions of Harm, Fairness, Loyalty, Authority and Purity, he and his team constructed more than 80 scenarios and filed dozens of questions to test in a systematic and quantitative way in what ways Americans judge the morality of machines different to how they judge humans. Finally, they fused all results into a moral space with three dimensions of morality — Harm, Intention and Wrongness — whereby discrepancies between the human and machine plain could be identified.
Clear becomes, we judge machines different to humans in both moral status and moral agency. In general, we have the tendency to like what machines do less than what humans do. Humans get more credit for success, while for AI responsibility shifts up in a hierarchy. However, we don’t accredit machines with intention and intentionality is the only factor that can worsen the judgement of human morality. Interestingly, with regards to labor displacement, Americans are more scared of foreigners than of robots.
“Intention is an enhancer of moral judgement.”
Moreover, there is strong variability within social groups. More educated people are more favorable towards machines, while women are more strongly against labor replacements of humans through machines.
If you are or are not yet struck by the relevance and beauty of the topic and findings → Buy his book that will be published soon!!!
What’s next?
- Read the literature and stay updated! Many of the keynote speakers will publish a book or article in the near future.
- Keep up with projects of the Young Scholar Initiative or get involved! Here also their engaging Facebook Page.
- Come to the next conference on the Geography of Innovation and Complexity in Mexico!
It was a great pleasure to meet you all and I hope to see you next year in Mexico! That’s it ;-)
