Mapping the production, diffusion and drivers of future technologies

Lead Research Organisation: University of Warwick
Department Name: Economics

Abstract

We propose to map and track the state of technological change in the UK, understand its drivers, impacts and help to improve the UK's productivity record via our collaboration and engagement with industry and policymakers. We focus on the role of frontier or 'future' technologies, such as AI, robotics, clean tech, blockchain and quantum. The contribution of these technologies to UK productivity will depend on the twin channels of their production and use (diffusion), the determinants of which we will seek to model theoretically and empirically.
We will build up quantitative evidence on economic activity in the UK's 'future-technology-producing' and 'future-technology-using' sectors. First, through description - which will include both sectoral and geographical elements; and second through building evidence on the extent to which financial or skills frictions constrain investment. We are therefore addressing two of the challenges set out by the PIN workshop: understanding and improving innovation diffusion; and understanding and improving regional and local productivities. An innovation in our approach is that we will use a range of emerging databases on technology-oriented firms linked to text-based information on their activities to build a comprehensive empirical picture of the future technology sectors. We will also use these databases to track the diffusion of technologies in novel ways, and plan to conduct a survey to shed more light on the barriers to diffusion, for SMEs in particular. We will work together with leading industry and policy stakeholders with expertise on new technologies, relevant data and techniques. Our empirical exercises will be guided by a conceptual framework that will model the effect of frictions in the financial market and the labor market on innovation, diffusion and economic growth.
Our project deliverables will consistent of four main components:
1. Future sector data and measurement network: Together with our project collaborators, we will host two major research meetings involving academics studying future technologies and relevant stakeholders from policy and business (e.g. NESTA and Be the Business). The first of these would seek to bring together the key stakeholders and set the agenda for research; and the second would involve presentations of our findings and developing an ongoing research agenda. We would also hold intermediate less formal working sessions.
2. Report: Our core deliverable after 18 months will be a report containing (i) a review of the relevant literature, empirical and theoretical; (ii) new descriptive statistics of the size, geography, structure and performance of the future technology sector in the UK (comparing and contrasting with traditional SIC based estimates); (iii) a conceptual framework to inform the drivers of the production and adoption of future technologies - by building a minimalistic model, the theory will aim to guide the empirical exercises using new data we collect or consolidate; and (iv) empirical modelling of production and adoption of future technology using new firm level databases and a survey, including external (to the firm) measures of the supply of finance/skills where possible.
3. Agenda for future academic research with new co-authorships from the project team: extending theoretical work and testing its predictions using our new data, and identifying areas where additional data are required. We envisage that the project will lead to the development of at least 2 academic papers, most likely one more focused on the FTP firms, and one on the FTU sectors, with paths to causal identification.
4. Data: Where the data we are using are purchased, we will seek to provide resources and guidance for researchers wishing to build on our analysis. According to standard academic practice, we would share our own generated survey data on an anonymised basis to interested parties, via the UKDS as appropriate.

Planned Impact

Here we set out who will benefit from our proposed research, and how.

Our work will be directly relevant for any policymakers involved in the design and evaluation of the UK's national industrial strategy and related policies, since the technologies of the future impact on and interact with all foundations and grand challenges in the UK's industrial strategy white paper. In addition to BEIS, this will include HMT, Cabinet Office, and also DfE and DCMS. The newly formed Industrial Strategy Council is currently developing metrics by which to evaluate the industrial strategy, and new sources of data and empirical relationships could be useful for informing their work. Members of our team are in regular contact with civil servants in these departments and we would seek to share work in progress and collaborate where possible. Moreover, our work will also be of relevance in the development of local industrial strategies and the GLA, is of our collaborators, are currently developing theirs. We have strong links with NESTA and IGL from our existing work, and we will seek to work closely with them in developing our ideas and approaches. Our work will also be of interest for international institutions such as the Organisation for Economic Cooperation and Development (OECD), who have begun working with some of these databases, and with whom our team have established links. Via our workshops, informal meetings and report, we hope to maximise the policy relevance and impact of our work.

Businesses are increasingly engaging with the productivity challenge - and this is relevant for the future producers and future users. Our collaborators Tech Nation and CognitionX have extensive contact with innovative firms that they support. CognitionX itself is a start-up in itself, providing an expert AI platform to other businesses, and disseminating knowledge via its annual "CogX festival". The GLA conducts SME outreach via its London Growth Hub, and will help ensure that our work remains policy relevant. Moreover, our research team has established links with "Be the Business", a new business community supported by the UK government, that is helping firms with the adoption of productivity enhancing technologies and business practices. Through these direct and indirect channels, our objective is that we can have real impact on businesses, growing the tech sector, improving diffusion and ultimately raising productivity and living standards.
Due to the scale of the productivity challenge and its implications for economic growth and the public finances, productivity is increasingly of interest to the wider public, and this is reflected by increased media coverage of the topic in recent years. We will therefore plan to reach the broader UK public mainly through writing blogs or articles and media appearances. The public can benefit from our work by gaining a better understanding of the technologies of the future, and how they are related to workforce skills. In our diffusion work, we will seek to communicate real world applications of future technologies to make them more tangible.

Finally, we seek to have impact on other academics working on these issues, through the publication of discussion papers and ultimately published articles. Via this project we will make a direct contribution to the work of the Productivity Insights Network, and also other groups working on similar issues, for example ESCOE and the What Works Centre for Local Economic Growth (the project team has strong pre-existing links with these groups). The project team is drawn from across institutions, and will help build a network of academics working on the technological change, innovation and diffusion of the technologies of the future. We will build on this through attending seminars or conferences and presenting work.

Publications

10 25 50
 
Description Key Findings

The project's key findings can be summarised according to three major areas that will be represented by working paper outputs and eventual academic publications. We are the final stages of writing up these working papers but provide links to slides in the "Other Outputs" section of Researchfish.


1/. The New Wave: Technology Diffusion in the UK during the 2010s.

Project Setting and Research Questions: As part of this research, we construct indicators of new technologies from the text of Burning Glass Technologies (BGT) job vacancy data for the UK, covering the period since 2012. Our indicators are constructed following the recent methodology of Bloom et al (2021) and are the first of their type for the UK. Specifically, we are able to build indicators for 29 technologies ranging from AI/machine learning, smart devices, electric vehicle technology (including batteries) and biotech.

However, AI/machine learning and cloud are our main focus in this paper since they are the most heavily diffused technologies present in the data. We summarise these as the core 'big data' technologies that were integrated into business processes during the 2010s.

The other main dataset we use is the Harte-Hanks (HH) ICT equipment database. This is a marketing intelligence database focused on ICT equipment measured at the establishment. Crucial, it is historical and we use it to build up a picture of ICT investments by UK firms in the early 2000s. Our key measure is PC intensity (i.e.: number of PCs per employee), which is a good summary measure of overall ICT investment.

Together, these dat sources give us information on technology adoption across 'two waves' of ICT

Using this setting, the key question that we address is: how skill-biased has the adoption of second-eave 'big data' technologies been compared to 'first wave' PC technologies? Alongside this, we also ask: do distinctions between types of skill matter for technology adoption? In particular, what is the role of STEM skills in technology adoption across the two waves?

Findings

We first present comprehensive information on the diffusion of 29 technologies over the 2010. After this, we present evidence at the region (Travel to Work Area (TTWA)) and then at the firm-level.
* In our first region level result, we confirm the result that PC adoption was strongly skill-biased during the ICT first wave in the late 1990s and early 2000s. Our skill measure is the ratio of degree to non-degree education individuals in an area and our specification is based on that of Beaudry, Doms and Lewis (2012) for the US. Indeed, our estimates are statistically very close to these US estimates, giving us confidence in our overall research design. These results are robust to the inclusion of regional controls such as population density and dummies for London / the South East.
* In our next result we examine skill bias for the second wave of big data technologies. We use two indicators of 'big data' technologies - cloud computing and ML/AI. This shows that second wave big data technology adoption has again been strongly skill-biased. When we benchmark the two waves against each other using standardised measures we find evidence cloud computing adoption is more strongly skill-biased than first wave PC adoption of ML/AI adoption.
* Building on this result, we then look at the relationship between technology adoption and STEM skills. This shows that region-level STEM skills are especially correlated with the the adoption of big data technologies. Indeed, STEM skills account for all of initial skills-technology correlation when it comes to second wave adoption.
* The next step of analysis involves firm-level data for second wave adoption. This is useful because it allows us to test for the effects of firm-specific skills. In line with the region-level results, STEM skills are strong correlates of big data technology adoption and this holds even with the inclusion of controls for firm labour force shares in other high skill occupations.
* Furthermore, this STEM-skills adoption effect is sharply increasing with the initial level of STEM skills. In turn, this points to a potential 'breaking away' effect whereby big data technologies are becoming concentrated within the most high tech technology firms.
* Finally, we make note of a potentially important trend or mechanism underpinning the adoption patterns that we are seeing. While human capital (defined in terms of degree education) has been increasing since the 1990s overall, STEM skill levels have actually been flat. Hence, while the adoption of second wave big data technologies is dependent on STEM skills the supply of these skills has not kept up with this implicit demand.
Exploitation Route We anticipate the following areas of research impact emerging from the project:
- The provision of a new database of technology indicators at the firm level as derived from job vacancy text data. In particular, this data gives specialised attention to trends in 'big data' technologies such as cloud computing and machine learning.
- A major contribution to the UK technology diffusion literature. In particular, we compare ICT-related technology diffusion for the PC-focused 'first wave' of the 1990s/early 2000s and then the second 'big data' wave of the 2010s. An important difference between the two waves is the higher STEM-skill bias of the second wave.



2/. Blitzscaling? The Changing Structure of UK Venture Capital

Project Setting and Research Questions: Over the last decade, there has been an international investment boom in relation to high technology firms with venture capital finance being a prominent part of this. We look at this for the UK using investment analytics data from the firm Beauhurst matched to a Companies House derived database of all new companies incorporated since the early 2000s.

As part of our database development we clarify the company accounts reporting rules in order to understand sample composition issues. This allows us to establish an overall sampling frame of the 'top tail' of new companies based on turnover reports. Practically, this also means that we can track firm growth patterns accurately, particularly fast growing firms.

Our main question relates the viability of the popular 'blitzscaling' hypothesis that has been a popular motivation for venture capital investment decisions in particular. The hypothesis holds that new technologies related to cloud and mobile computing have lower the fixed costs of firm investments as well as distribution costs. This has created opportunities for firms to quickly develop market power in new, consumer-facing markets. The classic examples given are typically Uber and AirB. In turn, this has led to a pattern of very large early investments in digitally-oriented markets that are intended to quickly build up market power.

The specific research question that we then address is then: do firm growth patterns since the 2000s support reflect the 'blitzscaling' hypothesis of new cohorts of fast-growing, technologically oriented firms? We study this by building a panel of start-up firms over time and relating their growth to

Findings
* UK investment changes driven by software sector and especially (bigger) later stage deals. Specifically, there has been no increase in average first round software deals over time but consistent increases in later round deals. This confirms the existence of the investment pattern associated with the 'blitzscaling' hypothesis.
* However, 'top tail' growth patterns have not changed. No distinctive change in growth or success rate for digital sector firms. There has been no change in the growth rate of successive cohorts at comparable stages of their life-cycles.
* Knowledge capital (patents, trademarks) is still a big, ongoing driver of 'extreme success', that is, fast growth as measured by the top decile of the sales distribution 5 years after co,many incorporation. But the relationship between knowledge capital and growth has not changed in a way that would directly support the 'blitzscaling' hypothesis.
* Finally, we also produce a new text-based taxonomy of firms. This does show increased differentiation within digital sectors in particular. But again, there is now associated shift in growth patterns that would support a strong 'blitzscaling' hypothesis.
Sectors Creative Economy,Government, Democracy and Justice,Manufacturing, including Industrial Biotechology,Other

 
Description We are in the final writing ups stages (some delays due to Covid mainly related to childcare responsibilities of key staff). But please see the following slide decks which summarise the findings and are the basis of the write. We will be rolling out dissemination during 2022. See the "Key Findings" section of researchfish for a detailed summary of results. https://warwick.ac.uk/fac/soc/economics/staff/mdraca/new_wave_2022_03_10.pdf https://warwick.ac.uk/fac/soc/economics/staff/mdraca/blitzscaling_tpi_2.pdf We have had many meetings with the Bank of England, BEIS and NESTA about the work. The project will create a major new body evidence on venture capital in the UK. This work has been presented internally and we are writing up with a view to release around May 2021, with accompanying online events. The Covid-19 pandemic did slow down some aspects of the project work (eg: extra childcare responsibilities) but we were able to re-organise.
First Year Of Impact 2020
Sector Construction,Creative Economy,Electronics,Government, Democracy and Justice,Manufacturing, including Industrial Biotechology
Impact Types Economic,Policy & public services