ECONOMIC COMPLEXITY ANALYSIS, Study Guides, Projects, Research of Public Policy

Download ECONOMIC COMPLEXITY ANALYSIS and more Study Guides, Projects, Research Public Policy in PDF only on Docsity!

A technical report for the research on

Innovation & Global Competitiveness

ECONOMIC

COMPLEXITY

ANALYSIS

March 2019

Authors: Penny Mealy and Diane Coyle, Bennett Institute for Public Policy,

University of Cambridge

The Bennett Institute for Public Policy at the University of Cambridge seeks

to rethink public policy in an era of turbulence and growing inequality. Its

research, teaching and policy engagement are guided by the need to devise

sustainable solutions to some of the most pressing problems of our time.

The views expressed in this report are those of the authors and, as usual, errors

and omissions in this report remain the responsibility of the authors alone.

The Panel commissioned studies in four areas, providing a thorough and

cutting edge analysis of key economic issues affecting the city region:

• Analysis of productivity, taking a deep-dive into labour productivity

performance across Greater Manchester (GM), including a granular

analysis of the ‘long tail’ of low-productivity firms and low pay;

• Analysis of education and skills transitions, reviewing the role of the

entire education and skills system and how individuals pass through key

transitions;

• Exploration of the city region’s innovation ecosystems, national and

international supply chains and trade linkages; and sources of global

competitiveness, building on the 2016 Science and Innovation Audit; and

• Work to review the infrastructure needs of Greater Manchester for

raising productivity, including the potential for new approaches to unlock

additional investment.

A call for evidence and international comparative analysis, developed

in collaboration with the Organisation for European Cooperation and

Development (OECD) and European Commission, also supported this work.

All of the Greater Manchester Independent Prosperity Review outputs are

available to download at www.gmprosperityreview.co.uk.

This technical report is one of a suite of Greater Manchester Independent

Prosperity Review Background Reports.

Contents

• 1. Introduction and scope...........................................................................................................
• 2. The Economic Complexity of UK Local Authorities and Industries
• 3. Economic Complexity of the GM Districts
  • 3.1 Mapping regional comparative advantage in the UK industry space
  • 3.2 Mapping Places in the UK Industry Space
• Technical Appendix
  • Calculating the ECI and PCI
  • Constructing the Industry Space
  • Proximity density (and distance)
• References

2. The Economic Complexity of UK Local Authorities and Industries The Economic Complexity Index (ECI) and Product Complexity Index (PCI) are measures of economic activity that have been shown to provide useful insights into the type of activities that distinguish prosperous from less prosperous places. Originally developed by Hausmann and Hidalgo (2009) to understand cross-country differences in productive capabilities from export data, the measures have since proved to be particularly successful at explaining variation in per capita GDP and predicting future growth rates across countries (Hausmann et al 2014a). Similar findings have been shown to apply to regional data (Gao & Zhao, 2018; Mealy et al 2018a, 2018b). In the regional setting, the ECI and PCI provide a useful way of understanding differences in local authorities’ industrial strengths. The ECI and PCI measures are calculated based on local authorities’ industrial strengths (see box). It is intuitive to think of the ECI and PCI as being related to the industrial diversity (the number of industries a particular local authority has a comparative advantage in) and the industry’s ubiquity (the number of local authorities that have an advantage in a given industry). However, a more accurate way to think about the ECI is as a measure that captures the most variation in the different dimensions of local authority industrial profiles. It provides a ranking that places local authorities with similar industrial profiles close together in the ordering, and local authorities with different industrial profiles far apart (Kemp- Benedict, 2014; Mealy et al 2018a). This ECI ranking is particularly interesting from an economic perspective because it is strongly correlated with UK local authorities’ earnings per capita (see Figure 1) and is significantly predictive of earnings growth (see the regression results reported in Table 1). The right hand side panel of Figure 1 highlights that local authorities with high ECI tend to be urban areas or cities, while local authorities with low ECI are more likely to be rural areas. Industrial strengths are measured using location quotients. Location quotients analyse the concentration of industrial employment in defined geographic areas. An industry 𝑗’s location quotient in a given area 𝑖 is calculated as the ratio of the industry’s share of employment in that location to its share of employment nationally. So if we define 𝐸!" as the number of people employed in industry 𝑗 in local authority 𝑖, then the location quotient for industry j in area i (denoted 𝐿𝑄!" ) is given by 𝐿𝑄!" =

𝐸!"/ ∑^ !𝐸!"

If a local authority has a location quotient greater than 1 (which indicates that the local authority’s employment share in that industry is the greater than the national average), the local authority is said to be ‘competitive’ or have ‘revealed comparative advantage’ (RCA) in that industry. Here we apply the Hausmann et al (2014a) algorithm to the location quotients to calculate the economic complexity metrics (ECI and PCI) for UK local authorities and industries.

Figure 1: Relationship between Local Authorities’ Economic Complexity and average annual earnings Table 1: Regression analysis of the relationship between growth in local authorities’ annual earnings and economic complexity Variables Annualised growth rate in average annual earnings (2011 – 2016) Economic Complexity in 2011 0.004*** (0.001) Log average annual earnings in 2011 - 0.053*** (0.005) Intercept 0.555*** (0.056) Observations 369 Adjusted R-squared 0. *** p-value <0.001, standard errors in parenthesis Workplace earnings data is sourced from the ONS Annual Survey of Hours and Earnings. Not all workplace earnings data was available for all 380 local authorities Table 2 shows the top and bottom ranked local authorities in terms of their ECI for the year

2015. High ECI ranked local authorities like the City of London, Tower Hamlets and Islington not surprisingly have similar industrial profiles to each other and different industrial profiles from the more rural areas like East Staffordshire, Sedgemoor, and Falkirk. Table 2: Top and bottom ranked Local Authorities by ECI ECI Rank Local Authority Rank Local Authority 1 City of London 371 Neath Port Talbot 2 Tower Hamlets 372 Pendle 3 Islington 373 Telford and Wrekin 4 Westminster 374 Rotherham 5 Southwark 375 South Derbyshire

East relate to finance, insurance, ITC and professional activities, while agricultural and manufacturing industries more likely to be concentrated in local authorities in Yorkshire and the Humber and Wales. Figure 2: Average ECI of Local Authorities in UK Regions Figure 3: Average PCI of SIC 3 digit industries for broader 2-digit SIC categories

3. Economic Complexity of the GM Districts Turning to Greater Manchester, Figure 4 shows the geographical distribution of local authorities’ ECI across the UK, with an inset highlighting the Greater Manchester city-region. Despite their relatively close geographic proximity, this inset highlights the stark differences between the GMCA boroughs. Manchester and Salford have the highest ECI, followed by Trafford and Stockport, which indicates they have relatively similar industrial profiles concentrated in higher-skilled service industries. In contrast, Wigan, Rochdale and Tameside have much lower ECI values, suggesting they have quite different areas of competitiveness, more concentrated in manufacturing activities. An effective industrial strategy needs to take account of these differences, as the realistic possibilities for future growth are likely to look very different across these different areas. Figure 4: Geographical Distribution of ECI across the UK

3 .1 Mapping regional comparative advantage in the UK industry space

To understand each GM district’s economic strengths and growth prospects in more detail this section turns to a networks-based lens. The UK Industry space is a network that helps visualise different types of industrial clusters that agglomerate together geographically. It represents industries as nodes linked to other nodes if they are more likely to cluster (be geographically co-located) within a local authority.

Figure 6: Manchester’s position in the UK Industry Space In Figure 7, we show Wigan’s position in the UK Industry Space. This allows us to see how different its strengths are from central Manchester. Wigan has few competitive strengths on the area of the UK industry space relating to research, finance and professional services. Instead it has a lot more employment concentrated in the centre of the network, in industries relating to construction, warehousing and storage, and wholesale and retail activities. Wigan also has areas of comparative advantage in a number of different types of manufacturing, such as textiles, chemicals, plastics and machinery. Textiles manufacturing Metals manufacturing Manufacturing of railway, military vehicles and optics Electrical product manufacturing Manufacturing of chemical and refined petroleum products Machinery manufacturing Call centers and social security support services Utilities Manufacture and repair of computers and consumer electronics Telecommunications, software programming and publishing activities Education Consumer goods manufacturing Water transport & ship building Television and radio broadcasting Fund management & insurance Professional services Agro-processing Research, data analysis & advertising Oil & gas extraction Business & management consultancy Industrial Concentration No Industrial Concentration

Figure 7: Wigan’s position in the UK Industry Space These differences in industrial structure are particularly important when considering industrial strategy. The UK Industry Space provides a useful visual tool for showing industries that local authorities are more likely to be able to build up in future given their present competitive strengths. Just as it is easier to make trousers if you already know how to make T-shirts, countries and regions are more likely to be able to diversify into products or industries that are ‘related’ (or require similar knowledge or inputs) to those they currently possess^3. So for example, since Manchester already has a number of competitive strengths in skilled professional services in the top left hand professional services area of the UK industry space, it would most likely find it easier to develop further industries in that cluster that can take advantage of the existing network of skilled professionals and knowledge already located in the city. In contrast, developing a research institute in Wigan may be more difficult – at least in the near-term – given Wigan’s current set of competitive strengths. However, manufacturing industries that can benefit from Wigan’s existing supply chains, storage and logistical capabilities could represent a more feasible development opportunity. In Figure 8, we show all 10 GMCA boroughs in the UK Industry Space. These plots provide a sense of the variation across industrial profiles within the city-region, and help give an indication of the types of industrial clusters that are present in each local authority. In the next section, we show how both the economic complexity measures and the network analysis presented here can be combined to identify potential future opportunities for industrial growth and development. (^3) See for example Hidalgo et al 2018 Textiles manufacturing Metals manufacturing Manufacturing of railway, military vehicles and optics Electrical product manufacturing Manufacturing of chemical and refined petroleum products Machinery manufacturing Call centers and social security support services Utilities Manufacture and repair of computers and consumer electronics Telecommunications, software programming and publishing activities Education Consumer goods manufacturing Water transport & ship building Television and radio broadcasting Fund management & insurance Professional Agro-processing services Research, data analysis & advertising Oil & gas extraction Business & management consultancy Industrial Concentration No Industrial Concentration

4. Identifying potential strategic opportunities for the ten Greater Manchester Boroughs Drawing on information about what UK local authorities are currently good at, we can identify new industrial opportunities that: (i) Are well aligned with the place’s current industrial strengths, and (ii) Have higher PCI, which could be advantageous in terms of growth and capability upgrading. To measure how well aligned a growth opportunity is with a place’s current industrial strengths, we consider a measure of ‘proximity density’.^4 By considering the probability that any two industries will be concentrated in a particular local authority, the proximity density metric captures the likelihood that a new industry could develop there, given its current industrial structure. So for example, if a place is already competitive in industries like accounting, tax consultancy and management consulting, its competitive strengths are likely to be more well-aligned or ‘proximate’ to the development of new industries such as insurance and fund management activities, and less well-aligned to say agro-processing or pulp and paper manufacturing.^5 Figure 9 shows the 10 GMCA boroughs. In these plots, green dots represent the local authority’s current industrial strengths, while grey dots are industries in which it is not yet competitive. The horizontal axis shows the distance (calculated as 1 minus proximity density) between a given industry and the local authority’s existing industrial strengths. The vertical axis plots each industry’s complexity (measured by PCI). The plot for Manchester city for instance (top left of Figure 9), shows a number of green industries in which it is already competitive including advertising, management consulting and computer programming. Industries shaded in purple represent new industrial possibilities that could be advantageous areas of competitiveness in the future. These industries, including market research and public opinion polling, trusts and fund management activities, and motion pictures, video and television, are not only well-aligned to Manchester’s current industrial strengths, they also have higher PCI. As discussed in section 1, higher PCI industries are concentrated in places with higher average earnings and growth performance. Such growth possibilities are sometimes referred to as ‘strategic bets’ (Hausmann et al 2014b). Although the probability of development in these areas is lower, their industrial success could stimulate significant future benefits in terms of greater diversification and growth opportunities in the longer term. Of course, given this greater risk of failure, promotion of these industries needs to be underpinned by careful feasibility analysis and a rigorous assessment of current binding constraints – such as availability of enough people with appropriate skills, or suitable space – that presently restrict development in these areas. A similar exercise can be carried out for the other Greater Manchester local authorities. Owing to its different existing set of capabilities, Stockport (top right plot of Figure 9) has a number of proximate opportunities that have low PCI, such as pre-primary education, landscape services, and residential care activities, but also higher some with PCI such as management consulting, software publishing and head-office activities. The plots for Wigan and Rochdale shown in the next two panels of Figure 9 both show a distinctly different pattern again. Wigan and Rochdale’s nearest future possibilities have low (^4) Originally developed by Hidalgo et al (2007) (^5) More information about the proximity density measure can be found in the Technical Appendix

PCI. For example, Wigan’s closest industrial opportunities include the sale, maintenance and repair of motor vehicles, repairing fabricated metal, machinery and equipment, and wholesale activities. Rochdale’s nearby industrial possibilities include manufacturing structural metal products and furniture, and construction activities. However, Wigan and Rochdale also have competitive strengths in a few industries that are more complex and less typical for their set of industrial capabilities. For example, Wigan has employment concentrations in office administration and business support service activities, while Rochdale has them in advertising, software publishing, and wireless telecommunication activities. The presence of these industrial concentrations could represent an opportunity for Wigan and Rochdale to build on these areas as a kernel of activity allowing them potentially to diversify beyond their traditional, low value, manufacturing-oriented industrial base. Bolton, Bury, Oldham and Tameside similarly have fairly industrial productive bases, with existing strengths and nearby growth opportunities tending to relate to less complex manufacturing activities. However, each of these local authorities also has a few key strengths in more complex, high-value areas such as management consultancy and telecommunications-related activities. Salford and Trafford have a more diverse portfolio of competitive strengths, with greater ability to leverage existing capabilities in market research, computer programming and financial services into more complex, higher skilled activities relating to data processing, information services, advertising and financial management. Figure 9: Identifying new industrial possibilities for the 10 GMCA boroughs Industrial Concentration No Industrial Concentration

Finally, it is important to emphasise that this analysis only represents an initial exploration of these places’ industrial strengths and future possibilities. Further analysis would need to analyse: (i) Whether efforts to encourage the development of a new area of activity in any specific location makes sense in terms of that sector’s broader growth prospects and demand profile; (ii) Whether there are binding constraints limiting growth in more complex areas of activity (such as skill shortages, lack of infrastructure or unfavourable regulatory environments) that policy will need to address if such industrial strategy policies building out from current specialisms to more complex activities are to succeed; (iii) The extent to which the activity is tradable and can serve markets beyond the local authority’s domestic demand. Tradable industries tend to have a stronger influence on a region’s growth and development, because unlike non-traded activities (such as barber shops, grocery stores, retail and other services, which tend to grow in proportion with the size of a local authority’s population), tradable industries are competing with other regions or overseas. As a result, tradable industries tend to have higher wage growth, higher productivity and patenting rates as they grow.^6 (^6) See for example Porter (2003)

• Market research and public opinion polling
• Management consulting activities
• Computer programming, consultancy and related activities
  • Manufacture of gas; distribution of gaseous fuels through mains
  • Freight rail transport
  • Transport via pipelines
    • Data processing, hosting and related activities
    • Leasing of intellectual property and similar products
    • Wholesale of information and communication equipment
    • ~Manufacturing of paints, varnishes and similar coatings
    • Manufacture of tubes, pipes, hollow profiles and related fittings of steel
• Reproduction of recorded media
• Telecommunication activities
• Wholesale and retail sail of information and communication equipment
• Architectural and engineering activities related to technical consultancy
• Specialised design activities
• Travel agency and tour operator activities
• Manufacture of other textiles
• • Manufacture of plastics productsTreatment and coating of metals
• Repair of fabricated metal products, machinery and equipment
• Materials recovery
• Printing and service activities related to printing

Technical Appendix Calculating the ECI and PCI The ECI and PCI are calculated on the basis of the following steps. First, we construct a binary 𝑀 matrix based on local authorities’ location quotients in different industries. In this 𝑀 matrix, rows represent local authorities, columns represent industries and 𝑀!" = 1 if local authority 𝑖 has an LQ in industry 𝑗 > 1 , and 𝑀!" = 0 otherwise. Summing across the rows of 𝑀 gives a local authority’s diversity (the number of industries it is competitive in), while summing across the columns of 𝑀 gives an industry’s ubiquity (the number of local authorities that it is concentrated within). Second, we calculate a local authority similarity matrix 𝑀, which is given by 𝑀 = 𝐷!!𝑀𝑈!!𝑀′, where 𝐷 is the diagonal matrix formed from the vector of local authority diversity values and 𝑈 is the diagonal matrix formed from the vector of product ubiquity values. The 𝑀 matrix captures how similar one local authorities’ industrial strengths are to another (see Mealy et al 2018a for more information on how to interpret this matrix). The ECI is defined as the eigenvector associated with the second-largest right eigenvalue of the matrix 𝑀. The PCI is symmetrically defined by transposing the binary 𝑀 matrix and finding the second largest right eigenvalue of an industry similarity matrix 𝑀, given by 𝑀 = 𝑈!!𝑀′𝐷!!𝑀. Constructing the Industry Space We construct the UK Industry Space from the binary 𝑀 matrix based on local authorities’ location quotients in different industries (defined above). Drawing on the methodology introduced by Hidalgo et al (2007), we first calculate the proximity 𝜙!" between two industries 𝑗 and 𝑘, which is based on their pairwise conditional probability of co-locating in a local authority and is given by 𝜙!" = min

!^ 𝑀!"

!^ 𝑀!"

Here we take the minimum of these terms to symmetrise the proximity measure and ensure 𝜙!" = 𝜙!". Two industries that have a very high proximity to each other have a very high probability of co-locating in a local authority, while two industries that have a low proximity to each other rarely appear in the same local authority. Two industries that have a very high proximity to each other have a very high probability of co-locating in a local authority, while two industries that have a low proximity to each other rarely appear in the same local authority. The UK industry space is essentially a network visualisation of this measure. Nodes in the industry space are industries, which are linked together on the basis of their proximity. However, if we were to visualise all links between all industries, it would be difficult to see any network structure. To create the network diagram shown in this report we follow Hidalgo et al’s (2007) approach and first construct the backbone of the network by calculating a maximum spanning tree from the 𝜙!" values. A maximum spanning tree of a given network