Clusters and the research opportunity of Big Data

Creating strong local economies and recognising the importance of place in achieving economic growth is at the core of the Local Industrial Strategies featured in the Government’s Industrial Strategy White Paper. The Local Industrial Strategies aim to identify local strength and challenges as well as unlocking opportunities to boost productivity.

Using clear evidence to identify and understand clusters of sectoral innovation activity will be guiding activities as part of the Local Industrial Strategies. The Smart Specialisation Hub is – as ever -fully dedicated to find evidence in data and analysis that help us identify ways to boost regional productivity.

In order to get a better grasp of the context for local strengths and challenges, we are committed to working closely with local partners like the City Region Economic and Development Institute (City REDI), based at the University of Birmingham, which develops an academic understanding of innovation in city regions. City REDI works globally, but the Institute supports the development of the West Midlands.

The report on Industrial Clusters in England, produced by City REDI in partnership with NIESR and SpazioDati, brings insights into how using Big Data can help understand local business assets and highlight local competitive advantages. Although the findings of the report reveal some relatively well known trends, it tests a data-driven methodology which has the potential to boost our insight into local trends, strengths and challenges.

The Hub supports the work done by City REDI and is looking forward to working with the Institute on future projects exploring the role of cities and place in increasing growth and productivity.

Clusters, and the research opportunity of Big Data – finding better ways to assess a specialised economy

By Dr Max Nathan and Rebecca Riley

BEIS Research Paper number 4

The policy expert looking to understand their local business assets is constantly constrained by current research tools and resources. Standard Industrial Classifications (SIC) provide a generic view of the world, as understood in the past. They don’t help us understand future sectors and clusters, they don’t help us understand cross sector embeddedness and linkages, and they don’t help us understand the local business ecosystem or its performance. This blog provides a summary of work carried out to test new techniques that might help us create a better understanding.

Last year, NIESR, SpazioDati and City REDI were commissioned by the then Department for Business, Innovation and Skills to generate new evidence on UK industrial clusters and to test the potentials and limitations of Big Data techniques applied to the study of the topic. Reflecting back on this work its findings are especially relevant when developing localised Industrial Strategies.

The report showcased an innovative data-driven approach to investigate the patterns of geographical clustering and functional integration across three sectors: digital health, financial services and the processing industry. These three sectors represent an emerging industry, an established service sector and a manufacturing sector with the presence of a formal cluster organisation. Although by no means covering the whole economy the work takes them as exemplars of the wider economy to test the new methodology.

Quantitative analyses were complemented by qualitative case studies based on interviews with key stakeholders. Semi-structured questionnaires generated detailed information on the evolution of the three selected clusters and the nature of the relationships between companies with other companies and local institutions.

It is concluded that general features of this Big Data methodology of industry classification can in general be applied to map clusters in emerging sectors not easily classified by the current Standard Industrial Classification system. However, a number of industry specific issues require ad-hoc solutions.

Quantitative Analysis

Internet data were used to identify companies belonging to each of the sectors being studied. Proprietary tools were used to screen and collect relevant information from companies’ websites, including geographical location, concepts describing a company’s activities and its web-links to other institutions.

An algorithm was applied to identify clusters based on the physical distance between companies. This approach allowed companies to be classified as part of the same sector even when they were classed with different SIC codes, and could identify clusters spreading over multiple discrete administrative areas.

This approach revealed similarities and differences in the patterns of geographical agglomeration across the three sectors. The largest urban areas emerged as important agglomeration areas for all three sectors. For example, London, Birmingham and Manchester were consistently identified as the largest sectoral agglomerations. Smaller urban areas had a different importance across sectors.

It is suggested that companies in these sectors are attracted to large metropolitan areas by factors that are common to the larger population of UK companies. These factors included the proximity with larger product and labour markets, and access to strategic tangible and intangible infrastructures within larger cities.

More stringent criteria to identify clusters were used to control for these factors. However, many clusters continued to be identified through this more stringent approach. This was interpreted as evidence of positive externalities from the co-location of similar companies within a specific geographical area.

The work demonstrated how important differences in functional relationships between companies and institutions between sectors may exist. Analysis of the network of web-links extracted from companies’ webpages found that digital health and processing industry companies’ websites contained frequent links to the sites of academic institutions. By contrast, financial services companies frequently linked to the same government websites common with other firms.

The study also identified the possible influence of sector-specific factors for digital-health. Oxford and Cambridge emerge as the only geographical areas where their relative concentration is at least two times the national average. No similar locations were identified for the remaining two sectors.

Qualitative Analysis

Qualitative data was used to gain insights into the strategic importance of the relationships developed by a company with other organisations inside and outside clusters. In addition, interviewers explored how companies’ officials perceived the value of locating within an industry cluster and their experience of the opportunities generated by geographical agglomeration and networking.

Some clusters are located where they are for historical reasons. The case studies on the financial services sector in Leeds and on the North East of England Process Industry Cluster (NEPIC) organisation in the Tees Valley emphasised the important role of historical legacy and central organisation in the establishment of these clusters. These two case studies shed light on the benefits arising from co-location of companies in the same industry or closely integrated industries.

The case study exploring digital health companies in Birmingham revealed that only six of the ten companies classified as digital health based on website data related to a strict definition of the sector. The other cases were generally pharmaceutical companies. In line with the limited number of inter-company web-links for this sector, the case study suggested that the agglomeration of digital health companies in Birmingham is not generally perceived as a functional cluster, and that there are no significant partnerships between the companies in the area.

All case studies confirmed the differing role of company-university relationships between sectors inferred by comparing network graphs of web-links. While companies in the NEPIC processing industry cluster and the digital health sector report strategic relationships with universities, this is not the case for financial companies.

Thoughts for the future

The work gives us a new insight into the limitations of the data and the importance of developing a holistic and varied approach to researching our business asset base. It demonstrates the importance of understanding the connections and value chains operating within a cluster not just what sector they belong to, and raises questions about intervention points for growth and whether they exist within the sector or within enabling companies.