032. What determines the geography of knowledge?

While it is obvious that a coal mine has to go where there is coal, or a fishing industry by the sea, there is no inherent natural geographic factor that affects the geography of knowledge. And yet there are very clear geographic patterns as to where these jobs locate. This section looks at why this is the case.

Why businesses locate where they do

An often cited benefit of locating in cities in the North is that they are cheaper places to do business than in the South. As Figure 6 shows, both workers and commercial space is cheaper in those cities that have replicated their economies. And yet certain businesses have chosen to set up in these more expensive locations.

To understand why this is the case, we need to understand why businesses make the location decisions that they do. Splitting businesses into three broad groups helps to explain these decisions:

High-skilled, knowledge-focused businesses

Because the main input of knowledge-focused businesses is knowledge and ideas, such businesses are reliant on high-skilled workers. This means that a city such as Bristol, where 39 per cent of residents have a degree, is more likely to attract a knowledge-focused business than Doncaster, where only 23 per cent of residents have a degree. As Figure 8 shows, there is a strong correlation between the two.7

But access to workers isn’t the only driver of location decisions for many of these firms – otherwise businesses would shun the high costs of central London for cheaper space on the M25, for example. The strength of the ‘knowledge network’ – the links between skilled workers – in a city is also an important consideration in firm location. The knowledge spillovers that are generated in these networks in strongly performing cities make knowledge-focused businesses more productive, more than offsetting the higher costs of doing business in such locations.

Knowledge-focused businesses are willing to pay a premium to access knowledge. The implication is that knowledge-focused cities compete primarily on their ability to give businesses access to knowledge, rather than on costs.

Figure 8: The correlation between skills and private sector KIBS jobs


Source: NOMIS 2015, Business Register of Employment Survey; NOMIS 2015, Annual Population Survey

Lower-skilled, more routinised businesses

Businesses that undertake more routinised activities, such as call centres and more traditional manufacturers, mainly require access to lower-skilled workers.8 And because they tend not to benefit from knowledge spillovers,9 there is little advantage to be gained from paying a premium to be based within a knowledge network.

Because these businesses are much more cost sensitive than knowledge-based businesses, the cost advantages that many northern cities have over southern cities make cities such as Barnsley and Wakefield a more attractive investment prospect to this type of enterprise. Indeed, this is why many companies split their head office location from their back office functions.10 So the cost benefits that the replicators offer does result in investment, but this investment principally comes from businesses that undertake more routinised activities. Box 3 discusses how this can be seen in patterns of inward investment.

Box 3: Patterns of foreign direct investment

Data on total investment by businesses in cities is not available. But by looking at the location of foreign-owned businesses, and the sectors they operate in, we can get an insight as to how attractive a city is to foreign investors.11

Figure 9 shows the correlation between the share of employment in low-knowledge industries in 1911 and the share of employment in foreign-owned businesses that is in manufacturing. If cities that fall into Quadrant B in Figure 9 are discounted to begin with, there is a positive relationship between the two indicators. Those cities that have historically had a high share of employment in low-knowledge industries have been more attractive to foreign manufacturing investment, which typically would prefer access to cheap land and low-skilled workers over the opportunity to benefit from knowledge spillovers in a knowledge network.12 The opposite trend is seen when looking at the share of jobs in foreign-owned businesses that are in KIBS activities.

Cities in Quadrant B have been more attractive to foreign manufacturers than their past industrial structure would suggest. Given that the majority of these cities are either current or former port or resort cities, this would suggest that these places have swapped logistics and tourism for manufacturing when the former were squeezed by containerisation and the rise of air travel. This aligns with data on the performance of manufacturing employment as a whole since 1911. Bournemouth, Blackpool, Grimsby, Southend and Worthing are five of just 11 cities where the absolute number of manufacturing jobs is higher today than it was in 1911.

Such patterns led to the creation of the term ‘branch plant’ economies in the 1970s and 1980s because of the growing separation between the knowledge functions of a company and its production functions. While inward investment in production facilities undoubtedly brings the benefit of jobs, the wider impact on a local economy in terms of the knowledge created by such investment is questionable, especially where there has been no historical specialism in the activity that the investment introduces.13

Of course, it is absolutely crucial to note that not all manufacturing employment is low-skilled, and this is likely to explain why Aldershot appears to be an anomaly compared to other Quadrant D cities. Despite being a more knowledge-focused economy, it has received a higher than expected amount of inward investment from manufacturing companies. Much of this investment is likely to have created high-skilled job opportunities – manufacturing employment in Aldershot is much higher-skilled than most other cities,14 and has a concentration of aerospace companies.15

Figure 9: Reinvention, replication and foreign investment


Source: Census 1911, ONS 2014, Business Structure Database

While such investment is invariably a good thing for the people who benefit from the jobs it creates, it serves to reinforce the replication of these economies. Such investment is likely to feel the pressures of globalisation most keenly in coming years too. While many northern cities are cheaper than southern cities, there are many locations around the world that offer an abundance of lower-skilled workers and land at an even cheaper price. Indeed, this is what has driven the offshoring of activities in the UK in recent decades.16

The opposite is the case for a knowledge-focused business. Although a software engineer costs three times as much in Silicon Valley as in Mumbai, most businesses do not transfer jobs to Mumbai because it isn’t just the skills of the engineer that are important, but also the proximity to a network of software and other digital businesses (and the tacit knowledge contained within them) which increases the productivity of that person.17 The presence and embeddedness of knowledge networks means that it is much easier to move a call centre or textile mill than a knowledge-focused business.

Locally focused businesses

While the first two types of businesses are likely to have regional, national or even international customers, businesses such as hairdressers and newsagents serve a very local market. Their location decisions are decided predominantly by where their customers live, work or trade from, and their fortunes are dependent on the performance of the first two types of business.

History matters – The ability to create and transfer knowledge is a long running trend

History has a large influence over the ability of cities to facilitate knowledge networks. Comparing jobs in knowledge services industries in 1911 and 2013, as is done in Figure 10, shows that those cities that had more knowledge-focused service economies in 1911 tend to have more knowledge-focused service economies today. For example, London, Reading and Bristol have historically been creators of knowledge, whereas Burnley, Huddersfield and Rochdale have not. Box 4 presents more detailed regression analysis on this finding.

Figure 10: Knowledge services jobs in 1911 and private sector KIBS jobs in 2013

Century-of-Cities-knowledge-jobs-1911- 2013

Source: Census 1911; ONS 2014, Business Register and Employment Survey

Box 4: The past as a predictor of the present

The industrial structure of a city 100 years ago is a good predictor of the share of KIBS jobs that are in a city today. Figure 10 shows the results of a regression of the share of KIBS jobs against 1911 industrial structure. There are three key findings:

Firstly, supporting the findings in Figure 5, those cities that were knowledge producers in 1911 tend to be so today, with the share of jobs in knowledge services in 1911 being a good predictor of the share of jobs in KIBS today.

Secondly, those cities that had more low-knowledge work in 1911 tend to have fewer KIBS jobs today. The share of jobs in extraction, dock working, light and heavy manufacturing and the share of workers classed as general labourers all have a negative and significant impact.

Thirdly, there is a big difference in the magnitude of these effects. The share of workers in general labouring and dock working have a much greater negative impact than those in extraction and manufacturing.

The regression also included a dummy for whether a city was a Victorian seaside town. A curiosity of the data is that seaside towns – including those that do not fall in the Primary Urban Area definition of a city, such as Eastbourne – tended to have large shares of workers in knowledge services in 1911. But as shown by Hastings, Southend and Birkenhead in Figure 7, and by the negative coefficient on the dummy variable, this is no longer the case.

Figure 11: Regression of share of KIBS jobs, 2013 against 1911 industrial structure


Dependent variable: KIBS jobs’ share of overall jobs in 2013. Regression run with robust standard errors.
* signigficant at the 10 per cent level ** significant at the 5 per cent level

Differing levels of tacit knowledge across cities in 1911 are likely to have caused these patterns. According to economic complexity theory,18 new industries don’t just appear out of nowhere – they evolve from industries that already have a presence in that country, with the tacit knowledge in the existing industry allowing the development of a new one. For example, it is easier to move from computer software development to smartphone app development, say, than it is to go from shirt production to app development.

Because tacit knowledge is best spread through face-to-face interactions, countries such as Bangladesh and Ghana cannot easily catch up to countries such as Japan and the USA. The same holds for cities within a country. This is why we have seen the long running divergence between cities such as Blackburn and Reading – the depth of the knowledge network is greater in the latter, which has supported the development of new industries as they have emerged.

The range of activities in each place is also likely to have a bearing on how cities develop. In Barnsley one in every two workers in 1911 was directly employed in mining. Mining is ‘less complex’ than many other industries and has few crossover skills with other industries. This has meant that there has been no natural development from mining into other more complex areas of activity in Barnsley once mining coal was no longer economically viable. The result is that the jobs that have been created in Barnsley have tended to be lower-skilled – 54 per cent of jobs in the city were lower-skilled jobs in 2013, the fourth highest of all British cities.19

Higher-knowledge businesses tend to have greater transferable skills with other industries, which makes the shift from software development to app development an easier move to make (what Jane Jacobs labelled as ‘related variety’20). This is one of the reasons why Bristol had the 8th highest share of higher-skilled jobs of all British cities in 2013.21 Box 5 uses IBM as an illustration to show how this evolution through innovation has occurred within one business.

Box 5: From punched cards to super computers – the progression of IBM

The history of International Business Machines (IBM) serves as a good illustration of how industries can develop.22 While the company is well known today for its computer software business, its current incarnation has evolved through a number of products over the last century.

IBM can be traced back to the merger of a number companies dealing in time management and punched card technologies in 1911. From here it produced a range of products ranging from time recorders to scales and meat slicers.

The company’s product line, through innovation, has become ever more sophisticated since. Calculation machines were followed by data storage and computer hardware development. This was followed by the personal computer, photocopiers and printers. The collection and management of data followed with IBM developing data encryption and barcode technology. And more recently this has evolved into computer services and software.

The IBM of today looks very different to the IBM of 1911. While examples of such businesses are rare (because those who don’t adapt ultimately go out of business), it illustrates the wider development that occurs in industries through time.

High-knowledge and low-knowledge pathways

By bringing these different ideas together – economic complexity, tacit knowledge and knowledge networks – we can set out two stylised pathways that (over the last 100 years) cities have developed along. The reinventors have followed a high-knowledge pathway, while the replicators have followed a low-knowledge pathway.

Figure 12 sets out both pathways. Those cities that had a high level of tacit knowledge in their economies in 1911 have been able to maintain and enhance the development of a knowledge network. This has made them attractive to new industries as they have grown, and it has meant that they have seen growth in new industries as older industries have matured and declined. And so their economies have been reinvented, becoming increasingly knowledge-focused.

That the replicators are still home to many thousands of jobs, and in many cases have increased the total number of jobs in their economies over the last 100 years, shows that there are still benefits to many businesses to clustering within these cities – they provide access to a large pool of lower-skilled workers to hire from. But this has meant that they have been less attractive to more knowledge-focused businesses, and so their jobs growth has tended to be in less knowledge-intensive, lower-paid work.

While helping to characterise the historical development of cities in Quadrants A and D, the pathways in Figure 12 say less about cities in Quadrants B and C. Cities such as Manchester and Leeds appear to have moved from a low-knowledge pathway to a high-knowledge pathway, while cities such as Southend have not capitalised on the stock of knowledge in their economies in 1911 and have shifted to the low-knowledge pathway.

Figure 12: High- and low-knowledge pathways


The ability of cities to switch pathways is likely to have been based on a change in the first two boxes of Figure 12 – that is a change in the levels of tacit knowledge and a change in the way that a city supports the growth of a knowledge network. The paper now turns to consider some of the elements that are likely to have influenced these changes.


  • 7 This does not imply causation – KIBS businesses are likely to locate where there are many skilled workers, and skilled workers are likely to be attracted to places with many skilled jobs.
  • 8 These workers may be specialised, but not necessarily highly qualified. Duranton G and Puga D (2001) Nursery Cities: Urban diversity, process innovation, and the life-cycle of products, London: CEPR
  • 9 Graham D (2007) Agglomeration Economies and Transport Investment, Journal of Transport Economics and Policy 41 (3)
  • 10 Leunig T & Overman H (2008), Spatial patterns of development and the British housing market, Oxford Review of Economic Policy, Volume 24, Number 1, 2008, pp 59-78
  • 11 We note that because of foreign buy-outs of UK businesses this can only act as proxy measure.
  • 12 Graham D (2007) ‘Agglomeration Economies and Transport Investment’in Journal of Transport Economics and Policy 41 (3)
  • 13 De Propis L and Driffield N (2006) ‘The importance of clusters for spillovers from foreign direct investment and technology sourcing’, Cambridge Journal of Economics 30 277-291; Phelps NA, Mackinnon D, Stone I & Braidford P (2003) ‘Embedding the multinationals: Institutions and the development of overseas manufacturing affiliates in Wales and North East England’, in Regional Studies 37 27-40
  • 14 In Aldershot there is one high-skilled manufacturing for every low-skilled job in manufacturing. For England and Wales, there are three low-skilled jobs in the sector for every high-skilled job. Source: ONS 2015, Census 2011.
  • 15 Aldershot has a location quotient of 3.0 for aerospace, defined as sectors 3030 and 3316. Source: ONS 2015, Business Register and Employment Survey
  • 16 We recognise here that other factors, such as transport costs, currency hedges and political stability also play a role.
  • 17 Moretti E (2012), The New Geography of Jobs, Boston: Houghton Mifflin Harcourt
  • 18 Hausmann R, Higaldo CA, Bustos S, Coscia M, Chung S, Jimines J, Simoes A, Yildirim MA (2013) The Atlas of Economic Complexity: Mapping Paths to Prosperity, Cambridge: MIT Press
  • 19 Defined as occupations in occupational groups Level 1 and 2 as classified in ONS (2010) Standard Occupational Classification 2010,Volume 1: Structure and descriptions of unit groups. Basingstoke: Palgrave Macmillan.
  • 20 Jacobs J (1969) The Economy of Cities, New York: Vintage
  • 21 Defined as occupations in occupational groups Level 3 and 4 as classified in ONS (2010) Standard Occupational Classification 2010, Volume 1: Structure and descriptions of unit groups. Basingstoke: Palgrave Macmillan.
  • 22 Pugh, EW (1995) Building IBM: Shaping and Industry and Its Technology, Cambridge: MIT Press