Policymakers have lately revived interest in encouraging clustering among innovative businesses. In last year’s Autumn Statement, the Chancellor announced his intention to refocus the Investment Zones programme to catalyse high potential clusters.1 As part of his report of the Commission on the UK’s Future, Gordon Brown identified clusters as an important part of Labour’s future economic growth strategy.2 Local authorities and pan-regional partnerships, such as the Midlands Engine, have similarly shown interest in understanding and supporting the clustering of knowledge-intensive activities.3
While research into clustering burgeoned in the 1990s, the roots of cluster theories lie in work on agglomeration which stretches back to the early 1900s.4 Clustering occurs because there are benefits to co-location – especially among complex, knowledge-intensive activities – which come from related firms sharing infrastructure and access to workers and credit.5 Co-location also allows the development of robust supply chains and knowledge spillover effects rooted in the exchange of knowledge within and between firms.6
Knowledge spillovers depend on face-to-face interactions. As a result, they require proximity and strong networks to operate most effectively.7 They are of great importance to startups and emerging high-technology businesses – the ‘new economy’ – for whom the sharing of ideas and techniques can be helpful for development. Presently in its infancy, the new economy is of national importance and is growing rapidly. Its constituent firms have thus far attracted billions of pounds of investment and Innovate UK grants.8 Recent evidence suggests that the knowledge spillover benefits enjoyed by these types of companies operate over very short, neighbourhood-level distances and weaken significantly after as little as 250 metres.9
Guided by these findings, this report examines the clustering of innovative new economy firms over very short distances. It does this by identifying hotspots where these valuable new economy firms cluster together (see Box 1 for the methodology) and by using statistical tools to understand why hotspots appear in specific locations.
Designed to find businesses that are deliberately positioned to benefit from agglomeration, the hotspot concept offers a perspective on clustering which emphasises the importance of proximity and place in supporting this nationally important innovative activity. In doing so, it sheds light on the factors which encourage the formation of hotspots in specific parts of cities and towns.
The report begins with an overview of the economic value of hotspots and their composition. Section 3 explores the geography of the clustered new economy. In Section 4, the report uses two types of regression analysis to assess the importance of inputs such as labour market size, skills, connectivity, universities, and large high-technology employers on hotspot locations and amounts of clustering in labour markets. Section 5 makes recommendations for policymakers.
Box 1: Data and clustering method
Centre for Cities’ research focuses on the UK’s 63 largest towns and cities. Unless otherwise stated, cities and large towns are defined as Primary Urban Areas (PUAs), using a measure of the built-up area of a large city or town, which sometimes spans beyond the core local authority. Full methodology is available at centreforcities.org/puas.
Cities are divided into two areas – city centres and suburbs – and the rest of the country is divided into hinterland and deep rural areas. City centres are defined based on the postcodes within a circle from the pre-determined city centre point. The radius depends on the size of the residential population (3.2km in London, 1.3km in cities with more than 550,000 residents, and 0.8km elsewhere). Suburbs are determined based on the postcodes that fall within the rest of a city.
Hinterlands are non-urban areas that are considered to be within commutable distance of cities. This varies from place to place and is determined by the average distance that a worker living outside a city travels to their job within it, defined using Census 2011 data. For example, the travel catchment area for London is 63km, but for Worthing it is 20km. The deep rural areas make up the remaining part of the physical landmass of Britain and fall outside of the travel catchment area of cities.
The data on new economy firms used in this report were collected by the Data City and were made available at the postcode level. The Data City uses ‘web-scraping’ of websites to identify companies engaged in cutting-edge activities and classify them according to a system of Real Time Industrial Classifications (RTICs). Specifically designed to group firms in emerging sectors, RTICs provide a more precise overview than that which can be gleaned from the official Standard Industrial Classification (SIC) used in most other research. Examples of RTICs include FinTech, advanced manufacturing, software as a service, and wearables.10
Centre for Cities uses 47 of 48 upper-level RTICs present in the Data City data (business support services were dropped for technical reasons). This leaves 88,162 new economy firms (NEFs) which together account for around 3 per cent of all businesses, according to the ONS in 2021. By examining their RTICs, new economy firms can be divided broadly into service and non-service activities. Service RTICs include FinTech and AdTech. Non-service RTICs include modular construction and advanced manufacturing.
There are two potential limitations of this dataset. First, because there are limited corresponding employment data, the analysis must treat all firms equally. In reality, some new economy firms will be bigger and more profitable than others. Second, the postcodes provided refer only to each firm’s registered address. This leads to the assumption that all the innovative activity undertaken by the firm happens in one location. Fortunately, sixty per cent of the businesses in the dataset have only one address, and firms with many addresses have been removed.
Accounting for instances where a registered address is a placeholder or the address of an accountant is more difficult. However, stress-testing of the data in previous Centre for Cities research suggested that the data were broadly robust to this problem.11 Cases where more than 500 businesses were registered to one postcode were manually checked to see whether they were the locations of accountants and, if so, a secondary address was used instead. Despite these problems, it is important to note that this dataset is the best available for analysis such as that contained in this report.
In order to identify and analyse locations in which new economy firms organically collocate over short distances, this report applies a clustering algorithm – Density-Based Clustering with Noise (DBSCAN) – to the postcodes in which new economy firms are located. DBSCAN is a popular algorithm used by economists, epidemiologists, and data scientists to identify clusters of various types.12
DBSCAN groups points which are close together and discounts those located in sparsely populated regions. Unlike other approaches to clustering, such as location quotients or specialisation indices, DBSCAN’s operation is not dependent on externally imposed geographies (such as wards or local authority districts). Using just two criteria – the minimum size a grouping can be and the maximum distance beyond which two points are not considered to be related – the algorithm is able to separate co-located observations from the rest.
Figure 1: An illustration of data grouped by DBSCAN
The hotspots identified by this report contain a minimum of 15 firms and are subject to a 250m maximum distance threshold. This means that any two new economy firms within 250m of one another are considered to be related by the algorithm. This conservative distance threshold increases the likelihood that the observed clustering is deliberate.
Various minimum size thresholds, ranging from 10 to 50 firms yielding between 25 and 800 hotspots respectively, were considered. Higher minimum size requirements resulted in hotspots located almost exclusively in city centres, and lower thresholds a more even balance across the country. London and the Greater South East stand at the centre of the clustered new economy in all specifications.
Defining clustering of any kind is always a difficult and somewhat arbitrary process. The agreed definition of clusters – ‘geographic concentrations of interconnected companies and institutions in a particular field’ – is broad and has no set rules regarding linkages and distance thresholds.13 The aim of the approach to clustering used in this report is to provide a precise picture of neighbourhood-level co-location in the new economy which is fair to different parts of the country and realistic in scope. The 344 hotspots identified in this report are comprised of 18,468 firms spread across every region.