Independent report

R&D skills supply and demand: long-term trends and workforce projections

Published 29 May 2025

© Crown copyright 2025

This publication is licensed under the terms of the Open Government Licence v3.0 except where otherwise stated. To view this licence, visit nationalarchives.gov.uk/doc/open-government-licence/version/3 or write to the Information Policy Team, The National Archives, Kew, London TW9 4DU, or email: psi@nationalarchives.gov.uk.

Where we have identified any third party copyright information you will need to obtain permission from the copyright holders concerned.

This publication is available at https://www.gov.uk/government/publications/rd-skills-supply-and-demand-workforce-trends-and-projections/rd-skills-supply-and-demand-long-term-trends-and-workforce-projections

Executive summary

Motivation

The Department for Science, Innovation and Technology (DSIT) aims to accelerate innovation, investment and productivity through world-class science, ensure that new and existing technologies are safely developed and deployed across the UK and drive forward a modern digital government for the benefit of its citizens.

Research and development (R&D) is fundamentally linked to a nation’s economic prosperity and its ability to respond to major challenges such as global warming, geopolitical tensions, and technological advancements. The R&D workforce is an important driver of innovation. This report adds to the evidence base for policies to enhance the inclusivity, diversity, and sustainability of the R&D workforce. These efforts will be crucial given the evolving needs for automation, artificial intelligence, and digitalisation.

The UK also faces considerable challenges in its labour market as a result of:

  • technological advancements
  • an ageing population
  • the on-going effects of the COVID-19 pandemic

The volume of skills shortage vacancies across all occupations in the UK more than doubled between 2017 and 2022, from 226,500 to 531,200 [footnote 1]. The government has announced that Skills England will build the evidence base needed to address skills gaps at a national and regional level [footnote 2]. Skills England will aim to build a high-skill, high-productivity workforce. This workforce will match employers’ needs and ensure that everyone, regardless of their background, can access the opportunities they need to thrive.

The following are needed to meet this demand:

  • a comprehensive understanding of workforce composition across regions
  • the demand for skills
  • the adequacy of the current labour force

There is limited empirical evidence on these issues specific to R&D occupations. This report addresses the current gaps and offers insights to address upcoming challenges.

Key findings:

In 2022, UK businesses spent £49.9 billion on R&D. £27.7 billion (56%) was spent on salaries for 652,000 FTEs R&D staff. Higher Education Statistics Agency (HESA) data from the higher education sector shows that in the 2023/24 academic year, 156,575 individuals were employed in research or teaching and research roles in Higher Education. 

This report has developed a cross-economy definition of the R&D occupations essential to R&D (see Section 8.2 for details). These occupations include:

  • science, research, engineering and technology professionals
  • higher education teaching professionals
  • selected business research professionals
  • technicians who are associated with these professions

Using this definition, analysis from this report shows:

  • According to the ONS’ Annual Population Survey (APS), there were an estimated 2.8 million people employed in occupations essential to R&D activities across all sectors in 2023 [footnote 3]. This has increased from 1.7 million in 2012 [footnote 4]. The percentage of the total UK labour force working in these occupations has risen since 2012.
  • Geographically, London experienced the highest increase in the proportion of the region’s total workforce being employed in R&D occupations between 2012 and 2021. In contrast, Wales and Yorkshire and the Humber consistently saw the lowest shares throughout the period of this study. Analysis in this report suggests these regional differences may be related to variations in the workforce qualification levels and the industrial make-up of these regions.
  • A few occupations make up a large proportion of overall workforce demand in R&D. For example, ‘Programmers and software development professionals’ and ‘IT business analysts, architects and systems designers’ vacancies comprised 42.4% of all R&D job postings from 2013-2022. Other factors, such as variable turnover rates, could also influence these figures.
  • The average hourly wages for individuals employed in R&D roles from 2012 to 2021 were substantially higher than those for non-R&D professionals (£21.28 compared to £14.56). However, wage growth for both groups remained relatively consistent during this period. When adjusted for inflation, real wages for R&D employees declined.
  • Analysis of job vacancies shows substantial variations in R&D salaries by occupation. IT/programming roles had the highest average salary at £55,000 per year in 2022, reflecting a rise since 2019. This is likely due to an increase in demand during the pandemic. Technician roles had the lowest advertised salaries in 2022 at around £31,000.
  • The Skills Imperative 2035 project made projections estimates in 2021. By applying these, we estimate that the number of R&D workers could reach 3.1 million under the ‘baseline’ scenario. This assumed the existing trajectory of technological advancement. The number of R&D workers could reach 3.5 million under the ‘technological opportunities’ or ‘human-centric’ scenario assumptions. These scenarios assumed accelerated technological adoption and increased investments in social care respectively.

R&D workforce characteristics

It is important to ensure talented individuals from all backgrounds have the opportunity to take part in R&D careers. This maximises the potential for R&D and fulfils the government’s mission to break down barriers to opportunity (Bell et al 2018)[footnote 5]. This ensures the UK can maximise the benefit of a skilled workforce by increasing participation in R&D from under-represented groups and increasing the overall number of people working in R&D.

  • The demographic characteristics of the R&D workforce in the UK differ from the wider UK workforce. Analysis of the Annual Population Survey data from 2012 to 2021 shows men were nearly 3 times more likely than women to work in R&D (8.8% of men vs 3.1% of women). This gender gap has persisted over time.

  • Notably, differences exist across various R&D roles. For instance, male programmers constitute 16.6% of the entire R&D workforce. This is nearly twice the 8.4% representation among female programmers. Conversely, higher education teaching professionals form 15.8% of women in R&D. This is almost 3 times the 5.5% of men. An even gender distribution within the workforce is linked to enhanced innovative performance in firms. It is also linked to a more diverse knowledge base and improved decision-making processes (Østergaard et al., 2011 [footnote 6]; Ritter-Hayashi et al., 2019 [footnote 7].

  • Gender earnings inequality continued even amongst the highly skilled R&D workers. Women earned less than men. This gap did not narrow significantly between 2012 and 2021. On average, the gender wage difference within the R&D workforce ranged between 9.8% and 19.5% over this time period in favour of men. The difference in average real hourly wages between genders in the non-R&D workforce ranged between 19.9% in 2020 and 21.9% in 2013.
  • Between 2012 and 2021, Bangladeshi, Black and Pakistani people were under-represented in the R&D workforce with 4.1%, 4.7% and 4.8% of the respective workforce populations working in occupations related to R&D. This is compared to 6% of the total workforce. This contrasts with Chinese (15.0%) and Indian (12.2%) people being more likely than the overall population to be employed in occupations related to R&D.
  • Workers’ level of education and the industry sector are 2 of the most powerful predictors of the likelihood of being in an R&D related occupation. R&D workers were more likely to hold high-level qualifications. Between 2012 and 2021, R&D workers were twice as likely to have an undergraduate degree (40.1% versus 21.3%). They were 12 times more likely to attain a doctorate degree (8.4% versus 0.7%). People working in sectors such as ‘information/communication’ (19.4%), ‘manufacturing’ (12.1%), and ‘professional/scientific/technical activities’ (11.7%) were among those most likely to work in R&D-related occupations.

Skill demand, shortages, and gaps in R&D

Currently, there is a lack of comprehensive evidence on the evolving skills demand within the R&D workforce and how well the workforce meets this demand. It is, therefore, important to develop a thorough understanding of the various skills and competencies in-demand for R&D roles and how these skills are likely to change in coming years. This understanding is vital to ensure the UK R&D workforce is well-equipped to face new challenges and demands. This report analysed the Employer Skills Survey and Lightcast job vacancies data to better understand skills demand in R&D.

  • In 2019, of the total 214,300 skills-shortage vacancies [footnote 8] in England, Wales and Northern Ireland [footnote 9], 15,600 (7.3%) were in R&D-intensive sectors [footnote 10]. By comparison, 5.4% of total vacancies were in R&D intensive sectors.
  • In 2019, across England, Wales and Northern Ireland skills-shortage vacancies made up 68% of all hard-to-fill vacancies. This proportion was much higher (92%) among R&D sectors. This highlights that employers faced greater challenges in recruiting people with appropriate skills, qualifications, and experience in R&D-intensive sectors.
  • In 2019, establishments (a branch or a site of an organisation with at least 2 staff on payroll) in R&D-intensive sectors with skills shortage vacancies found it difficult to source applicants with technical/practical competencies. These competencies include specialised skills or knowledge, complex problem-solving abilities, and advanced or specialist IT skills.
  • Certain soft skills were also difficult to find among applicants in R&D intensive sectors with skills shortage vacancies. These skills included time management and task prioritisation, customer handling skills, and team working.
  • Certain regions showed higher proportions of hard-to-fill and skills-shortage vacancies in R&D-intensive sectors. For instance, in the East Midlands’ R&D sector, hard-to-fill and skills-shortage vacancies represented 65% and 63% of all vacancies, respectively. This reflected greater recruitment challenges compared to London, which had the lowest proportion (23% hard-to-fill and 22% skills-shortage vacancies).

Qualitative evidence

Twenty-two semi-structured interviews were conducted with R&D employers and employees. These interviews provided valuable insights that enhance other important findings from the project. Insights from these interviews included:

  • To increase equality, diversity and inclusion in the workforce, interviewees reported the importance of role models, alternative pathways into STEM, and improving employment practices. For example, they highlighted the value of using inclusive language in job adverts and facilitating flexible working arrangements.
  • Interviewees also identified potential barriers to attracting and retaining a skilled and diverse workforce. They identified a low number of women and ethnic minority applicants for STEM-related jobs due to low numbers of these groups taking relevant courses at school and university, the loss of potential STEM workers to higher paying industries, such as banking and finance, and the limited visibility and recognition of technicians in R&D.
  • Interviews also highlighted employer and employee perspectives on high-value skills in R&D. They showed digital, data science, and cyber skills are increasingly valued due to new technologies and vast amounts of data. Interviewees also recognised the importance of interpersonal skills, communication and leadership in R&D roles.

1. Introduction

Innovation is a critical driver of a country’s economic growth and prosperity (Hasan and Tucci, 2010 [footnote 11]; [footnote 12]. It is also considered the foundation of responsive actions trying to address current economic disruptions and substantial challenges, such as:

  • global warming
  • geopolitical tensions
  • the ageing population
  • the increased technological needs related to automation, artificial intelligence (AI), and digitalisation

The UK labour market will evolve drastically in the next decade. As a result, a clear understanding of the relevant trends and challenges is needed. Examples of these trends and challenges include:

  • technological advancements
  • an ageing population
  • labour market inequalities
  • environmental imperatives (Taylor et al., 2022 [footnote 13]; Green et al., 2021 [footnote 14]

These challenges are intensifying in the aftermath of the COVID-19 pandemic and the ongoing labour and skills shortages.

These trends can heavily influence R&D. However, there has been limited empirical evidence considering how the structure and characteristics of the R&D workforce vary across the UK regions and nations (Cardenas Rubio and Hogarth, 2021 [footnote 15]). Similarly, little is known about whether the proportion of R&D workers in the total labour force has changed over time. This becomes more important in light of the recent shocks and structural changes the COVID-19 pandemic caused to the UK economy (De Lyon and Dhingra, 2021 [footnote 16]. There is some evidence that the pandemic led to disproportionate regional effects on employment and the labour force composition and an increase in shortages of qualified workers (ONS, 2021a [footnote 17]. It is therefore important to have a better understanding of the composition and characteristics of the R&D workforce, skills gaps, and determinants influencing this.

Currently, there is a lack of comprehensive evidence specifically around the evolving skills demand within the R&D workforce. There is also a lack of evidence of how well the labour force meets this demand. It is important to develop a thorough understanding of this and how these in-demand skills are likely to change in coming years. This understanding will ensure the UK R&D workforce is well-equipped to face new challenges and demands.

This report aims to provide an in-depth analysis of trends in the composition of the R&D workforce, including personal characteristics and occupational makeup of the workforce. We employed a varied approach, using diverse data sources and a range of methodologies. Employer skills demand is explored, with an investigation of ‘hard-to-fill vacancies’ and ‘skills shortage vacancies’. There is also a detailed look at how the skills required in R&D job postings have changed over time. Detailed insights into the skills required for the R&D workforce, perspectives on the influence of equality, diversity and inclusion (EDI) issues and other topics are explored via in-depth interviews with employers and employees in the R&D sectors. Furthermore, the report looks into future projections of R&D employment. It analyses anticipated shifts in employment levels for R&D-related occupations and their relative proportion in overall employment under various future scenarios. This work also examines how the outcomes differ across different regions and nations of the UK.

2. Understanding skills matches and mismatches, employers’ and workers’ perspectives: A literature review

2.1 Overview and general findings

This section provides a selective review of the literature on aspects of skills demand and supply and their interactions.

Skills refer to the ability to use knowledge to achieve objectives and are central to a holistic concept of competency. Skills can be difficult to define. In fact, classifications vary in the literature.

The OECD (2019) [footnote 18] categorises skills into:

  • cognitive
  • social and emotional
  • practical and physical types

The Industrial Strategy Council (2019) [footnote 19] classification includes:

  • formal qualifications
  • subject-matter expertise
  • workplace skills

Another commonly used term, ‘technical skills’, relates to specific job requirements.

Accurately measuring skills can be challenging due to their multi-dimensional nature. Typically, proxy measures like highest educational attainment, are used for measurement, but these conventional proxies are limited in that they do not offer a clear representation of skill levels (Dickerson and Morris, 2019 [footnote 20].

The table below outlines some fundamental concepts of skills mismatches. Section 10.1 provides a more detailed description and discussion of these concepts based on the relevant literature.

Main concepts of skills mismatch/equilibrium

Concept Description
Skills mismatch Occurs when skills supply does not meet skills demand.
Skills shortage Occurs when supply is less than demand.
Skills surplus Occurs when supply is greater than demand.
Skills gap Occurs on the internal labour market when managers assess that employees do not meet the competence levels required for a particular job role.
Undereducation Occurs when the education level of workers and others in the available talent pool is less than is required.
Overeducation Occurs when the education level of workers and others in the available talent pool is greater than is required.
Skills underutilisation Occurs when workers are employed below their skills level.
High skills equilibrium Occurs when the economy demands high level skills in high wage, high productivity ‘good jobs’ and the supply of labour meets demand.
Low skills equilibrium Occurs when demand for skills is low and so there is a lack of incentives for employers and individuals to invest in skills.
Skills deficit Occurs when the supply and demand for skills are at a sub-optimal (or feasible) level.

Workforce mobility between sectors, occupations, or geographical regions can serve as a potential tool in addressing skills gaps. Some sectors have been found to form ‘skills basins’, where workers circulate and spread knowledge, supporting learning and innovation (O’Clery and Kinsella, 2022 [footnote 21]). Geographically, both international and internal mobility offer the potential to bridge skill gaps. While internal geographic mobility in the UK appears to be witnessing a decline (Champion and Shuttleworth, 2017 [footnote 22]; Green, 2018 [footnote 23]), immigration can introduce the required skills into the workforce. The immigration regime is an important consideration that can influence how employers can benefit from migrant labour. For example, this includes factors such as whether migrants are authorised to work based on a specific employer demand (in the form of a job offer) or are simply admitted based on their characteristics (Sumption, 2019 [footnote 24]).

Training and upskilling remain central for addressing skills gaps. Various forms of training include:

  • on-the-job or off-the-job
  • formal or informal
  • certified or non-certified

There appears to be a notable decrease in participation rates and expenditure on these types of training in the UK. However, online or e-training is experiencing an uptick (Taylor and Green, 2021 [footnote 25]). Employers can be reluctant to train, in part due to concerns around poaching (Green et al., 2020 [footnote 26]).

An emerging trend is the rise of micro-credentials [footnote 27]. They offer a promising way to upskill and address immediate skill shortages. They have gained popularity for their cost-effectiveness, quick delivery, and alignment to specific employer or industry needs. A workplace with employees with complementary qualifications helps achieve a skills match at the firm and economy level.

2.2 Employers’ perspectives

Important skills gaps exist within the current workforce, including:

  • time management
  • leadership skills
  • customer skills
  • analytical skills and digital skills (Green and Taylor, 2020 [footnote 28])

Digitalisation and COVID-19 have extended skills gaps. Current workers need to develop new skills. For example, the emergence of AI, automation, and digital technologies (Allas et al., 2020 [footnote 29]; Dondi et al., 2021 [footnote 30]) could lead to severe skills shortages in basic digital, core management, and science, technology, engineering, and mathematics (STEM) skills. Previous estimates suggested that by 2030, approximately 7 million additional workers could be under-skilled for their job compared to 2019. Over 10 million people could be under-skilled in leadership, communication and decision-making (Industrial Strategy Council, 2019 [footnote 31]). This is particularly relevant as employers appear to place increasing value on staff having a mix of social and behavioural skills (Green and Taylor, 2020).

Further new skills needs are emerging due to the government’s net zero by 2050 target. These skills relate to supporting innovation and diffusion of zero-carbon technologies and ensuring a ‘just transition’ for individuals displaced by changing skills needs (Li et al., 2020 [footnote 32]). Work Foundation research with low carbon businesses in Lancashire found skills shortages are limiting growth and development in the sector. This research suggests considerable change is needed in the skills system to respond to increased employer demand (Walker, 2020 [footnote 33]). To support this transition, Nesta has identified 4 skills priorities:

  • a consistent government-led data-driven approach to upskilling
  • new funding and expanded wider support (for example, childcare) to enable individuals to receive training through non-traditional education providers and establish new-industry non-traditional training programmes
  • adopt a mission-oriented approach to skills policy
  • support inclusive innovation in the green economy through regional clusters and upskilling (Kapetaniou and McIvor, 2020 [footnote 34])

How employers can best address skills gaps remains the question. There is a strong economic case for ‘upskilling and reskilling’. Doing so leads to positive economic returns to employers in 75% of cases. It also generally contributes to a productivity uplift of 6% to 12% (Allas et al., 2020 [footnote 35])Strategies that employers can adopt to effectively encourage reskilling include:

  • conducting strategic workforce planning to identify skills gaps within organisations
  • developing employee training options
  • promoting a culture of lifelong learning (Allas et al., 2020)

‘Refining the skills system’ and improving opportunities for learners to gain relevant work experience are suggested ways of helping close skills gaps (for example, the apprenticeship system). Surveys and interviews with employers in the Engineering Construction Industry revealed challenges facing apprenticeship recruitment, such as:

  • a lack of suitable work
  • no current need for apprentices
  • preference to hire graduates or experienced staff over apprentices (ECITB, 2020 [footnote 36])

Reforming the Apprenticeship Levy could create a more flexible training levy. Such reform would address employer confusion regarding the range of roles apprenticeships can be used for and differences in how apprenticeships operate across the devolved nations (CIPD, 2021 [footnote 37]. PhD graduates could also play an important role in addressing skills issues. Recruiting PhD graduates “sustains and enhances the mission and growth of the organization” (McAlpine and Inouye, 2021 [footnote 38]). This suggests developing tailored pathways from PhD studies into non-academic employment could be helpful.

2.3 Workers’ perspectives

Employers, universities, and government bodies have shown increased interest and concerns about shortages of highly skilled workers in the UK (for example, BEIS, 2017 [footnote 39]; Industrial Strategy Council, 2019). Science, Technology, Engineering and Mathematics are strong correlates of R&D and innovation activities. A report by the University of Cambridge shows that the rate of increase of people studying STEM courses at the undergraduate level was nearly twice as large as students enrolling in non-STEM disciplines over the period 2010 to 2018 (Policy Links, 2021 [footnote 40]). This represented 47% of total undergraduates in the academic year 2017/18. In the UK, the proportion of STEM graduates was higher (44%) than in comparator countries, such as France and the US (37%). Despite this, the Cambridge report reveals that the share of R&D workers (defined as ‘researchers’, ‘technicians’, and ‘other supporting staff’) was lower in the UK labour force compared to comparator countries (France, Germany, Switzerland, and Korea). Smith and White (2018) [footnote 41] argue that this is likely due to graduates’ occupational decisions and employers’ differing hiring procedures across sectors. The authors find that only one-third of STEM graduates are employed in ‘highly skilled (HS) STEM jobs’ 6 months after completing their studies, with variation by discipline (for example, 16% for biological science graduates).

Academic literature has well-established that the cultural, ethnic, and birthplace diversity of a workforce generally has a positive influence on a country’s:

Similarly, a balanced gender makeup of the workforce is correlated with increased innovative performance of firms, resulting from a diversified knowledge base and more effective decision making (Østergaard et al., 2011 [footnote 49]; Ritter-Hayashi et al., 2019 [footnote 50]).

However, the British Science Association (BSA) (2021) [footnote 51] reports that women are under-represented in the STEM labour force in the UK. They covered 27% of total STEM workers in 2019 (relative to 52% of the rest workforce). Furthermore, there is evidence that women holding STEM qualifications are more likely than men to work in ‘caring’ occupations but less likely to make their way up the career ladder by accessing managerial jobs (White and Smith, 2021 [footnote 52]). Specific ethnic minorities (Black, Pakistani, and Bangladeshi workers) are also underrepresented in STEM-pertinent activities. However, ethnic diversity is more pronounced in certain industries, such as the health sector (BSA, 2021). Though earnings for STEM degrees are higher than for arts, humanities, and education (Britton et al., 2016 [footnote 53]; Walker and Zhu, 2018 [footnote 54]), ‘ethnic penalties’ to earnings persist even after accounting for the subject studied (Zwysen and Longhi, 2018 [footnote 55]). Finally, people with disabilities are less likely to be employed in STEM (11%) than other sectors (14%), with a higher gap for women with disabilities (BSA, 2021).

COVID-19 pandemic saw working from home become more normalised. This shift could somewhat alter the extent to which the socio-demographic composition of local areas affects the under-representation of employees from disadvantaged backgrounds in the STEM workforce. However, the existing literature emphasises that the lack of comprehensive and consistent over time data does not allow a complete picture of the demographic characteristics of the STEM labour force across UK regions, disciplines, sectors, and occupations to be painted (BSA, 2021).

Schneider and Sting (2020) [footnote 56] performed semi-structured interviews with employees working in manufacturing firms in Germany. The purpose of these interviews was to explore employees’ perceptions and attitudes towards the fourth industrial revolution (‘Industry 4.0’). Schneider and Sting’s primary aim was to understand how workers react to the digitalisation-induced change. The authors find varied patterns in the level to which employees accept this change and identify 5 cognitive frames that measure workers’ perspectives:

  • ‘utilitarian’
  • ‘functional’
  • ‘anthropocentric’
  • ‘traditional’
  • ‘playful’

Their study shows that managers are more likely to use the functional frame, whilst their employees adopt a mixed frameset.

Finally, there is evidence that participating in further training raises job satisfaction. However, the effects are smaller among employees with a disability (Pagán-Rodríguez, 2014 [footnote 57]). Hence, upskilling the current workforce and designing targeted training programmes can address skills shortages and improve firms’ productivity. The latter is correlated with employee job satisfaction (Böckerman and Ilmakunnas, 2012 [footnote 58]).

This section draws on data from the UK Annual Population Survey to better our understanding of trends in workers’ likelihood of being employed in R&D occupations and depict the profile of the R&D personnel. The primary objectives are to:

  • Explore differences in workers’ likelihood of being employed in occupations related to R&D across UK regions and over time
  • Investigate differences in participation rates in R&D employment by socio-demographic characteristics, level of education, health status and industry sector
  • Adopt econometric techniques to model the average probability of being employed in an occupation related to R&D conditional on the above mentioned factors

3.1 Initial findings

The purpose of this section is to:

  • present the occupational distribution of R&D employment across genders and regions and nations
  • use various demographics to calculate the proportion of people in R&D-related occupations
  • show how R&D employment trends have evolved over recent years

We identify particular subgroups within the labour force that are under-represented in R&D occupations. The aim is to inform the relevant policy debate and interventions that would lessen existing inequalities. While workforce numbers are not a perfect reflection of demand for R&D labour and skills, they serve as a useful indicator.

3.2 Size of the R&D workforce

This report has developed a cross-economy definition of the R&D occupations essential to R&D (referred to as the ‘core’ definition, see Section 8.2 for details). These occupations include:

  • science, research, engineering and technology professionals
  • higher education teaching professionals
  • selected business research professionals
  • technicians who are associated with these professions

There were an estimated 2.8 million people working in occupations essential to R&D in 2023. This is an increase from an estimated 1.7 million in 2012. The numbers have increased each year with the exception of 2021, when the number of workers fell slightly before increasing again in 2022, as illustrated in Figure 1a.

Figure 1a. Estimated number of people employed in all core R&D occupations in the UK over time

Source: ONS ‘Employment in research and development occupations by nationality, selected years 2009 to 2023’[footnote 59] and ‘Employment in research occupations and development professions by ITL2 area, UK, 2004 to 2020[footnote 60]

The APS datasets are weighted by the ONS to reflect the size and composition of the general population using the most up-to-date official population data. For more information see the Annual population survey (APS) QMI[footnote 61]

Table 1: Estimated number of people employed in all core R&D occupations in the UK, (weighted, based on SOC 2010)

Year Number of R&D workers
2012 1,685,220
2013 1,746,319
2014 1,828,493
2015 1,926,604
2016 1,957,109
2017 2,050,784
2018 2,129,510
2019 2,257,232
2020 2,390,655
2021 2,332,971
2022 2,487,509
2023 2,799,632

Source: ONS APS ‘Employment in research and development occupations by nationality, selected years 2009 to 2023’ and ‘Employment in research occupations and development professions by ITL2 area, UK, 2004 to 2020’ (weighted data).

3.3 Alternative sources of data on parts of the R&D workforce

The above estimates use data from the ONS Annual Population Survey to estimate the number of people employed in occupations essential to R&D across all industries.

Higher Education Institutions employ a large number of staff vital to R&D. They record staff numbers via the HESA. In the 2023/2024 academic year, there were 246,930 people employed in Higher Education Institutions in the UK.

  • 88,725 were ‘teaching only’ staff
  • 106,290 were ‘both teaching and research’ staff
  • 50,285 were ‘research only’ staff

The number of people employed in ‘teaching only’ roles has increased at a faster rate than those employed in research or teaching and research roles.

Figure 1b: Number of people employed by UK Higher Education by role

Source: HESA, HE academic staff by ethnicity and academic employment function 2019/20 to 2023/2024[footnote 62].

3.4 Distribution of R&D workers by occupation

The remaining analysis in this section uses unweighted pooled data from the Annual Population Survey from 2012 to 2021 to enable detailed breakdowns by maximising sample sizes(see methodology Section 8.4 for further details). This section outlines the key findings, with detailed breakdowns available in the accompanying data tables and annex data tables ODS files.

We first look at occupations of the R&D workforce and the occupational makeup and gender breakdown of the R&D workforce between 2012 and 2020 (based on the ‘core’ definition described in Section 8.2). The distribution of workers across the occupations that form the R&D workforce varies significantly by gender. Among the R&D occupations, ‘programmers & software development professionals’ represent the largest group (14.6% of all R&D workers in the UK), followed by ‘IT & telecommunications professionals not elsewhere classified (n.e.c.)[footnote 63]’ (9.1%) and ‘higher education teaching professionals’ (8.1%). The proportion of male programmers is twice as large as that of women (16.6% of men in the R&D workforce are programmers versus 8.4% of women in the R&D workforce). In contrast, the share of women teaching in higher education is 3 times bigger than that of men (15.8% versus 5.5%).

The results also reveal gender disparities in the engineering professions favouring men. The differences in proportions range from 1.7 percentage points (for electronics engineers) to 5.1 percentage points (for engineering technicians). Previous evidence suggests that, in their school years, women put a higher value than men on courses relevant to language, reading, and health (Jerrim and Schoon, 2014[footnote 64]). As a result, these preferences likely influence women’s future choices to take non-STEM educational and occupational paths. However, they are equally qualified to men to access STEM jobs, such as engineering. On the contrary, the share of ‘biological scientists & biochemists’ and ‘laboratory technicians’ is substantially higher among women than men. The proportion gap stands at 7.3 and 6.7 percentage points, respectively. (Further breakdowns are available in Table 2 of the accompanying data tables.)

Qualitative evidence – relevant findings from interviews with R&D employers and employees

Interviewees discussed how underrepresented groups, in particular young girls, can have less awareness of, and fewer aspirations to work in, STEM careers. This appears to be linked to having fewer role models and fewer discussions about STEM careers with family/friends, as well as cultural values. Participants cited their desire to learn through practical experience as their primary motivation for pursuing a STEM pathway. For instance, a technology transfer engineer revealed that this tendency for the practical side of things was clear throughout her educational journey. Her personality, including her natural curiosity and eagerness to learn how to solve problems, influenced her decision to take a practical route and study engineering despite coming from a non-engineering family. However, according to a higher education institution (HEI), senior technician, parents can sometimes have conflicting expectations because of cultural reasons.

“You can probably imagine I come from a traditional Asian background. My parents wanted me to become a pharmacist or a doctor or a lawyer kind of thing. So, I did the traditional A-levels, which was biology, chemistry, etc. Needless to say, I didn’t do too well initially but I retook them. And during that phase of retaking my A levels, I sat down and said to myself ‘I don’t think I want to go down the traditional university route of taking a degree’. I wanted an experiential sort of way of doing it because I learn better by experience. So, I undertook an apprenticeship. As you can imagine, my parents, being quite traditional Asian parents, weren’t very pleased with that. … Well, after the 2 years of this apprenticeship, I undertook a degree with [name of university] and the rest is history.” (HEI senior technician 2)

Example of good practice: embedding school role models in engineering projects

A civil engineer employee described how she actively engaged with local schools when leading on projects in local areas. Engagement was designed to raise awareness of engineering and its importance in society as well as the diversity of the engineering workforce:

“We’ve gone in and done a whole school assembly, and then we’ve taken children out to do tree planting on a flood risk project site. And in that example, we were hoping to gain awareness. So, I want to show children a little bit about what engineering is, the importance of engineers to our society quite broadly. And that I’m a woman in engineering. And so, this is partly what an engineer looks like. It’s not just one face, there’s a lot of different faces within engineering. And I think in understanding the concept of flooding, and climate and the environment, are all topics that I’m really keen to promote as important to children.” (Civil engineer employee)

The distribution of these occupations also varies regionally. For example, a higher than the UK average share of R&D workers in London make up occupations related to information and technology, such as ‘programmers & software development professionals (Table A1 in the accompanying Annex data tables). In contrast, engineers and technicians represent a smaller percentage in the capital relative to the UK average. These regional differences in the makeup of R&D employment likely reflect differences in the opportunities offered across industries and the demand for specific innovation skills.

There is also variation linked to the level of education with 72.7% of R&D personnel holding an ‘NQF level 4 and above’ qualification compared to 39.8% of non-R&D workers. (See Tables A2 and A3 in the accompanying Annex tables. See Section 10.5 for a definition of NQF levels). R&D workers are also twice as likely to gain a first degree (40.1% versus 21.3%) and 12 times more likely to attain a doctorate degree (8.4% versus 0.7%). This reflects the typically higher skill level of R&D workers, as they invest more years in education.

3.5 R&D workforce participation by region and individual characteristics

Between 2012 and 2021, men were far more likely than women to work in R&D-related occupations. The UK average share of R&D workers in the total labour force over this time was at 8.8% and 3.1%, respectively, for both genders (Table 3 in the accompanying data tables). The West Midlands sees the lowest proportion of women in the R&D workforce (2.5% of women working in the West Midlands). This disparity likely reflects the industrial landscape in the West Midlands. This landscape sees a notable presence of manufacturing and engineering sectors, such as automotive and aerospace. These sectors traditionally attract more men than women to the workforce. In contrast, participation of women is markedly higher in London (4.4%).

The South East shows the highest share of R&D workers (7.6% – Figure 1c). However, this region sees a relatively large gender gap in this R&D participation rate (7.4 percentage points), followed by the South West (6.7 percentage points). On the other hand, the lowest shares are observed in Wales (5.0%), Yorkshire & the Humber (5.1%), and the North East (5.4%). This picture likely reflects the regional imbalances in the distribution of R&D investments (Forth and Jones, 2020[footnote 65]). It also reflects differences in the industry structure across regions, which shape the demand for R&D skills.

Figure 1c. Proportion of the total workforce in each region working in R&D occupations between 2012 and 2021 (‘core’ definition)

Note: Differences between regions of less than 0.4 percentage points are not statistically significant. Source: APS pooled datasets (July 2012 to June 2021) unweighted, authors’ calculations

The average likelihood of being an R&D worker gradually increases by age until the late 30s. It reaches its highest value at 7.5% for people aged between 31 and 40 years (see Table A4 in the accompanying Annex data tables). This picture is reversed for older workers. The proportion of workers under 25 years employed in R&D occupations is relatively small. Young people can still be in some form of higher education before embarking on R&D employment. Moreover, our definition of the R&D workforce only includes workers classified in the top 3 major groups of the Standard Occupational Classification. This implies that it can take a few years for young workers to climb the career ladder and secure a managerial/professional job.

There are also substantial differences in ethnic representation in the R&D sectors across regions.

  • Chinese (15.0% of the Chinese workforce) and Indian (12.2%) people are more likely than others to be engaged in R&D activities. Specifically, the proportion of Chinese people in R&D roles in the West Midlands and the East of England (19.3%) is the highest in the UK.
  • Scotland (20.6%) and the South East (19.4%) see the largest proportion of Indian people working in R&D professions.
  • On the contrary, Bangladeshi (4.1%), Black (4.7%) and Pakistani (4.8%) workers are the least likely to be employed in R&D-related jobs. (See Table 4 in the accompanying data tables).

The above findings may be linked to ethnic differences in cultural capital. They may also be linked to the value attached to specific subjects of study and professions (for example those relating to sciences and engineering), with subsequent consequences for individuals’ occupational choices. There is also variation in employment in R&D by nationality, with people from Eastern Europe least likely (3.6%) to be part of the R&D workforce[footnote 66] (see Table A5 in the accompanying Annex data tables).

Table A6 in the accompanying Annex data tables shows that of the total workers with health conditions/illnesses lasting over one year, 5.4% are employed in R&D-related professions. This is lower than the share of workers with no health issues (6.2%). The gap in the likelihood of conducting R&D between these 2 groups of workers is more prominent in Scotland (1.3 percentage points) and the North East (1.1 percentage points). People with learning difficulties[footnote 67] are less likely than others to be employed in R&D-related jobs. Specifically, during the period 2012 to 2021, the UK average proportion of people with learning difficulties in the ‘core’ R&D workforce was 2.8%. This is significantly lower than that of people with no learning difficulties (6.0%)[footnote 68].

Qualitative evidence – relevant findings from interviews with R&D employers and employees

When discussing EDI issues, interviews revealed that employers tended to concentrate on the fact that there are not enough women or people from ethnic minority backgrounds applying for STEM jobs, primarily due to the low number of young people studying for relevant courses at school and university.

“First of all, you need a diverse group of people to apply for jobs in the first place. So, if you have got girls at school who don’t study STEM, they’re not going to go to university and do technical degrees that we would recruit them from.” (Large engineering company director)

The majority of interviewees stressed the importance of developing better systems to promote role models and raise awareness of STEM careers in schools. This would address the lack of diversity in specific R&D occupations (such as engineering).

“I think people won’t necessarily apply for manufacturing jobs if they can’t see their own ethnic group there. I think that might be quite an issue.” (STEM asset representative)

“My understanding is that a lot of decisions are already made for women in STEM, sort of at primary school age, and really quite young.” (Civil engineer employee)

Based on the unweighted APS data from 2012 to June 2021, the share of R&D workers in the total workforce (based on the ‘core’ definition) rose from 5.4% in 2012 to 7.9% by June 2021 (see Table A7 in the accompanying Annex data tables). The increase during this time period was particularly clear during 2020 and the first half of 2021 and could be attributed to 3 reasons.

  • The COVID-19 pandemic shock may have increased the demand for particular skills/jobs related to R&D. More businesses adopted automation and digital technology to adapt to the new challenges and trends, such as remote working and e-commerce (World Economic Forum, 2020[footnote 69]; Enterprise Research Centre, 2022[footnote 70]).
  • The negative influence of the pandemic on unemployment was disproportionately distributed across occupations and sectors. Lower-skilled professions were the hardest hit (ONS, 2021a[footnote 71]). Early evidence from several countries suggests that online job advertisements declined considerably during the first months of the pandemic (OECD, 2021[footnote 72]). Nevertheless, the pandemic has not equally affected the demand across levels of education and industries. For example, as the OECD (2021) policy brief mentions, the number of job postings relating to lower education levels (secondary and below) dropped by 40% in the UK after March 2020. It decreased by 30% for high-skilled workers (attaining a postgraduate or doctoral degree). Similarly, leisure and hospitality were the most affected sectors. According to our SOC definition, the R&D workforce is composed of high-skilled workers only. These workers are less likely to transition into unemployment. This means the share of R&D employment in the total labour force increased during the pandemic.
  • The shift from the SOC 2010 to SOC 2020 in the APS data may have affected, to a lesser extent, the figures for the first half of 2021 (January to June) relative to the previous years. However, we did apply the same rules to select the R&D-related codes across both SOC 2010 and SOC 2020.

As Figure 2 shows, the gender gap in the participation rates in R&D employment has not narrowed during 2012-2021. Men remain significantly more likely than women to work in R&D-related professions throughout the period of our study.

Figure 2. Proportion of the workforce in R&D occupations in the UK (‘core’ definition) by gender and year

Source: APS datasets (July 2012 to June 2021) unweighted, authors’ calculations

London witnessed the most significant growth in the share of R&D employment in the total labour force between 2012 and 2021 (4 percentage points), followed by the North West (3.8 percentage points) and Northern Ireland (3.3 percentage points)[footnote 73]. In contrast, the East of England (0.9 percentage points) and Yorkshire & the Humber (1.4 percentage points) saw the slowest growth figures in the same period. Workers residing in the South East had the highest chances of being involved in R&D activities until 2020 (Table A7 in the accompanying Annex tables). However, for the first time, London was ranked top amongst all the UK regions in 2021. Some of this regional growth may be correlated with changes in the proportion of R&D workers by occupation. For example, ‘engineering professionals not elsewhere classified (n.e.c.)’, ‘IT business analysts, architects & systems designers’ and ‘programmers & software development professionals’ showed the largest absolute increase in their proportions between 2012 and 2020 (see Table A8 in the accompanying annex data tables). The latter 2 occupation groups see a notably higher share in London than in other UK regions (Table A1 in the accompanying Annex data tables).

3.7 Wage growth analysis

Investigating the wage growth of R&D occupations relative to the supply of R&D workers and the wage changes in non-R&D jobs may provide insights into how the demand for innovation skills evolves over time and whether skills mismatches occur (Green F., 2016[footnote 74]. According to this approach, if the relative supply of R&D personnel grows (for example, an increasing proportion of STEM graduates) while the average salaries of R&D employees remain stable or increase, there is evidence that the demand for R&D workers/skills has risen at least to the same degree as the relevant supply. The exact rate of increase depends on the so-called ‘elasticity of substitution’ in production. The latter term refers to how easily businesses can switch between different factors of production (like R&D labour versus technology) when costs change. The hourly wages of R&D workers (adjusted for inflation) showed only small fluctuations from 2012 to 2021. Specifically, the UK average hourly wage ranged from £21.95 in 2012 to £21.46 in 2021 (see Table A9 in the accompanying Annex data tables). This occurred despite the significant increase in the relative supply of highly educated workers (particularly STEM graduates) by universities and other institutions in the same period (Cardenas Rubio and Hogarth, 2021[footnote 75]). R&D workers (regardless of gender) earn more than their non-R&D counterparts. The average difference in wages across all years of our analysis stood at £6.72 per hour. Comparatively, the wage premium of R&D workers is higher in the South East (£7.27) and London (£7.07), but lower in Northern Ireland (£5.39) and the North East (£5.75). However, the UK average gap in real hourly wages between R&D and non-R&D workers appears to have narrowed over the last few years, decreasing to £5.78 in 2021 from £7.29 in 2015.

From the above discussion, it is difficult to firmly assess whether there are skills mismatches in the R&D sector using the wage growth method. The approach has disadvantages when concentrating on short-term analysis, especially if the signals linking the supply and demand in the labour market are weak (Gambin et al., 2016[footnote 76]). There can also still be significant skills shortages in subgroups of individuals (‘shortage hotspots’). The large number of professions that compose the R&D workforce can conceal differences between R&D occupations. Therefore, other approaches, such as analysing job posting data, could be another way to evaluate how the demand for R&D occupations/skills varies across (and within) regions and evolves over time (see Section 5).

Finally, our results suggest that the well-documented gender earnings inequality persists even amongst the highly skilled R&D workers. This gap has not narrowed significantly over time (Figure 3). The gender wage difference within the R&D workforce ranged, on average, from 9.8% in 2018 to 19.5% in 2016. Notwithstanding, the gender pay gaps in R&D employment are generally lower than those in the rest of the labour force. In particular, the difference in average real hourly wages between genders in the non-R&D workforce varied between 19.9% in 2020 and 21.9% in 2013.

Figure 3. Average real hourly wages in the UK by gender and ‘R&D worker’ status

(‘core’ definition) (2012 to 2021)

Source: APS datasets (July 2012 to June 2021) unweighted, authors’ calculations

4. Who is more likely to work in R&D? An econometric insight

While descriptive evidence is a good beginning, it does not always capture the complete context. Econometric models give us a more comprehensive insight into how various factors interact. With these models, we can measure the influence of elements such as region, gender, ethnicity and other characteristics on the likelihood of being an R&D worker. This is done while ensuring other important variables are also factored into the analysis.

This section presents the relationship between the probability of being employed in an occupation related to R&D and all the independent variables contained in the logistic regression. To achieve this, we use the average marginal effects (AMEs) and the average adjusted predictions (AAPs) described in Section 10.4.

In this econometric approach, we gradually introduce specific factors into our analysis. We then observe how they influence the relationships we established. By progressively adding more variables into the regression models, we see how these variables modify or refine the observed links between worker characteristics and the likelihood of being in an R&D role. The findings reveal that the relationship between the chief variables of interest and the likelihood of conducting R&D does not change drastically compared to the descriptive evidence provided in Section 3. This remains true even after accounting for several factors in the regression models

The strongest predictors of the probability of working in R&D-related jobs are:

  • gender
  • ethnicity
  • nationality
  • the level of highest qualification held
  • the industry section of the workers’ main job
  • the workplace region

Specifically, as Table A10 in the accompanying Annex data tables shows, men are more likely than women to be employed in R&D-related occupations. This is even after allowing for differences in a wide range of other characteristics. In particular, the gender difference in the likelihood of conducting R&D ranges from 5.9 percentage points in the baseline regression (Model 1) to 4.6 percentage points in the extended specification (Model 2) and the regression with interaction terms (Model 3)[footnote 77].

Chinese and Indian people are more likely than others to work in R&D occupations. This aligns with the descriptive results presented earlier. More specifically, when we factor in other variables, the difference between these 2 ethnic groups and the White reference category is 5.0 and 1.4 percentage points, respectively (see Model 2 of Table A10 in the accompanying Annex tables).

Being married/in a civil partnership is positively associated with the probability of being employed in R&D occupations. This might be linked to other unobserved characteristics[footnote 78]. For example, economic motivation might influence married people to a greater degree to enter highly skilled employment.

Conversely, having dependent children lowers the probability of working in R&D occupations. This could be partially connected with the ‘motherhood penalty’ that affects career advancement in high-salaried jobs (Waldfogel, 1998[footnote 79]; Viitanen, 2012[footnote 80].

The level of education also plays a crucial role in the probability of being employed in an R&D role. More precisely, workers holding ‘level 4 and above’ qualifications are far more likely to enter R&D-related occupations. The difference ranges from 4.6 percentage points (for those educated at ‘NQF level 3’) to 8.0 percentage points (for employees with no formal qualifications).

In addition, we find no evidence that workers’ long-standing health issues influence the probability of being employed in an R&D occupation after accounting for the other explanatory variables.

The direction (sign) of the effect the workplace region has on the dependent variable changes once we move from the baseline model to Models 2 and 3. This deserves further attention. According to Model 1, people working in most UK regions (except for the South East) are less likely to be employed in R&D-related professions than those working in London, keeping all else fixed. For instance, the relevant difference between London and the West Midlands is 1.1 percentage points.

However, this picture is reversed in Model 2. This model encompasses further control variables referring to education and job/industry characteristics. The likelihood of participating in the R&D workforce now becomes lower in London than in the rest of the UK regions, though the differences remain relatively small. This suggests that once accounting for differences in the workers’ level of education and their industry sector, the regional imbalances are minimised, and the London ‘premium’ vanishes. Indeed, according to our data, the proportion of people whose highest qualification is ‘NQF level 4 and above’ is significantly higher in London than elsewhere in the UK. Likewise, the share of workers employed in industries related to ‘Information/Communication’ and ‘Professional/Scientific/ Technical activities’ is considerably higher in London. In several cases, this reaches to 2 times the respective shares in other UK regions. As Table A10 in the accompanying Annex data tables shows, these variables (the education level and the 2 industry sections mentioned above) are strongly and positively associated with the likelihood of conducting R&D. This implies that the effect of London on the dependent variable was overestimated in Model 1 (that is, before we adjusted the analysis to account for workers’ level of education and industry sector).

Workplace size also corresponds with the chances of accessing R&D professions. Workers employed in medium/large-sized firms are more likely to be employed in R&D occupations. This result ties in well with the findings from the UK Innovation Survey (UKIS). These findings show that:

  • 50% of large companies were innovation active[footnote 81] between 2020 and 2022 compared to 36% of small and medium-sized enterprises (DBT, 2024[footnote 82])
  • Self-employed people are on average 3.5 percentage points less likely to work in R&D professions than employees (Model 1 of Table A10 in the accompanying Annex tables)
  • Public sector workers have a higher chance of 2.0 percentage points to work in R&D occupations than those in the private sector (Model 2). The probability of working in R&D gradually increased between 2012 and 2021. For instance, in 2021, the average likelihood of working in R&D professions was 1.7 percentage points higher than in 2012 (based on the extended model’s estimates)

Figures 4 to 5 below and Figures A2 to A5 in section 9 illustrate the predicted probabilities of working in R&D occupations by:

  • gender
  • ethnicity
  • nationality
  • level of highest qualification
  • workplace region
  • industry section in the main job

These probabilities are obtained from Model 2 of the logistic regression (post-estimation). In summary, the average adjusted likelihood of engaging in R&D activities is higher for:

Figure 4. Probability of being an R&D worker (‘core’ definition) by level of highest qualification held (average adjusted predictions)

Note: The graph shows the predicted likelihood of being an R&D worker conditional on the variables included in the extended logistic regression model (Model 2 of Table A10 in the accompanying Annex data tables). The shaded areas depict the 95% confidence intervals of the average adjusted likelihood. Source: APS pooled datasets (July 2012 to June 2021), authors’ calculations.

Figure 5. Probability of being an R&D worker (‘core’ definition) by industry section in main job (average adjusted predictions)

Note: The graph shows the predicted likelihood of being an R&D worker conditional on the variables included in the extended logistic regression model (Model 2 of Table A10 in the accompanying Annex data tables). The shaded areas depict the 95% confidence intervals of the average adjusted likelihood. It should be noted that the ‘Mining/Quarrying’ and ‘Electricity/Gas supply’ sectors account for a small proportion of total employment relative to other industries, which might contribute to greater variability and less precision in the regression estimates for these sectors.

Source: APS pooled datasets (July 2012 to June 2021), authors’ calculations.

5. Lightcast job vacancy analysis and insights from the Employer Skills Survey 

This section of the report sets out important findings from the analysis on demand trends by occupation for the R&D workforce.

Looking at demand for all R&D occupations as a percentage of total demand, job postings for ‘core’ R&D occupations made up 15.3% of all job postings in 2022. This figure is higher at 23.6% when broader R&D occupations are included. This brings the total demand for all R&D occupations to just under a quarter of all demand for all occupations in 2022. Between 2013 and 2022, the percentage of job postings for R&D occupations in the Lightcast dataset decreased. The share of non-R&D job postings rose by 2.6 percentage points (pp). Between 2019 and 2021, there was an increase of around 1.0-1.5 pp in the proportion of core R&D roles in all job postings. This reflects the COVID-19 pandemic’s potential influence on the acceleration of digitalisation in 2020 and 2021. Subsequently, the share of R&D roles saw a slight decline in 2022. This might be tied to a general economic recovery across all sectors (see Table 5 in accompanying data tables).

Table 6 presents the demand for each individual ‘core’ R&D occupation as a percentage of total demand for ‘core’ R&D occupations. It also presents the corresponding value for the percentage of the current R&D workforce according to the analysis of Annual Population Survey (APS) data presented earlier (see Section 4). Table 7 in the accompanying data tables shows how this demand has changed over time for groups of these occupations.

Table 6: Proportional demand and workforce composition in R&D occupations (‘core’ definition)

SOC 2010 Occupation title Lightcast % APS % Difference (pp)
2136 Programmers and software development professionals 28.6% 14.6% 14.0
2135 IT business analysts, architects and systems designers 13.8% 5.9% 7.9
3119 Science, engineering and production technicians n.e.c. 4.7% 2.2% 2.5
3113 Engineering technicians 6.8% 5.1% 1.7
2461 Quality control and planning engineers 2.4% 1.8% 0.6
3116 Planning, process and production technicians 1.8% 1.4% 0.4
2129 Engineering professionals n.e.c. 6.0% 5.7% 0.3
2123 Electrical engineers 2.7% 2.5% 0.2
2121 Civil engineers 4.2% 4.0% 0.2
2122 Mechanical engineers 4.0% 4.0% 0.0
3115 Quality assurance technicians 1.7% 1.7% 0.0
2124 Electronics engineers 1.0% 1.6% -0.6
2111 Chemical scientists 0.8% 1.4% -0.6
2114 Social and humanities scientists 0.4% 1.0% -0.6
2429 Business, research and administrative professionals n.e.c. 1.7% 2.4% -0.7
3114 Building and civil engineering technicians 0.4% 1.1% -0.7
2425 Actuaries, economists and statisticians 1.0% 1.8% -0.8
2113 Physical scientists 0.4% 1.2% -0.8
2426 Business and related research professionals 1.1% 2.1% -1.0
3112 Electrical and electronics technicians 0.5% 1.5% -1.0
2127 Production and process engineers 1.3% 2.6% -1.3
2126 Design and development engineers 2.0% 3.7% -1.7
2150 Research and development managers 0.4% 2.4% -2.0
2119 Natural and social science professionals n.e.c. 0.3% 2.5% -2.2
2112 Biological scientists and biochemists 2.0% 4.3% -2.3
2139 Information technology and telecommunications professionals n.e.c. 6.1% 9.1% -3.0
3111 Laboratory technicians 0.8% 4.2% -3.4
2311 Higher education teaching professionals 3.4% 8.1% -4.7
Total Total 100% 100% Not applicable

Note: The table compares the demand for R&D occupations (as indicated by job postings data from Lightcast) against the composition of the R&D workforce (based on Annual Population Survey data). Lightcast data is from 2013 to 2022. APS data spans 2012 to 2020. The table is ordered by the largest percentage differences in demand versus workforce composition. Source: Analysis of Lightcast and APS (unweighted) microdata.

The analysis reveals that the percentage of demand for certain R&D occupations in the Lightcast dataset differs notably from the percentage of the R&D population for each occupation in the current labour force. The most notable difference is in ‘Programmers and software development professionals’ and ‘IT business analysts, architects and systems designers’. Here, the proportion of demand is higher than the proportion of the R&D population in these fields by 14.0 pp and 7.9 pp, respectively. These 2 occupation groups make up 42.4% of all R&D job postings over the given timeframe.

Overall, demand for most occupations in the Lightcast dataset is lower than their corresponding APS values, particularly for ‘Higher education (HE) teaching professionals’, and ‘Laboratory technicians’. This suggests that a small number of occupations make up the majority of demand for R&D occupations. For some occupations, this could be the case. However, these differences could be explained by a range of other factors:

  • potentially higher turnover rates
  • shorter terms of employment for certain occupations
  • higher participation in online job postings when compared with other methods of hiring for certain occupations

For example, R&D managers may be more likely to be promoted from within an organisation and stay in their role for longer. Other occupations, such as programmers and software development professionals, may frequently use online job postings for job searching and be more likely to switch between roles frequently. LinkedIn’s own research from 2021 to 2022 showed that IT and business consulting, as well as tech organisations, exhibit some of the highest turnover rates across all organisations[footnote 83]. This may be a contributory factor to the relatively high proportion of demand occupied by IT and programming roles.

While comparing Lightcast and APS data offers valuable insights, it is important to recognise their limitations.

  • The non-overlapping years (2021 to 2022) in the Lightcast dataset may leave out recent developments in R&D occupations that the APS data available was unable to capture at the time of the analysis
  • The move to online job advertising likely influences how certain roles are represented in Lightcast data. This risks introducing a potential sampling bias. This shift also suggests that the proportions of job postings might not accurately reflect true demand trends. This is because advertising practices may have changed from 2013 to 2022
  • The ‘demand’ in Lightcast job postings and workforce representation in APS data could reflect different market dynamics. A higher workforce proportion but lower demand in postings could indicate a role’s saturation or decline. The opposite scenario might suggest an emerging or underserved field

As a result of these nuances, it is important to consider both datasets cautiously when understanding the R&D job market.

We are not only interested in the trends in workforce demand over time. We are also interested in how demand for different R&D occupations varies between the different regions across the UK. To this end, we analyse workforce demand across the 9 English regions – East of England, East Midlands, Greater London, North East, North West, South East, South West, West Midlands, and Yorkshire and The Humber. We also analyse other nations of the UK – Wales, Scotland and Northern Ireland.

The Greater South East (GSE) is a grouping of 3 English regions – East of England, Greater London and the South East. Demand for R&D workers varies greatly between regions of the UK.

  • Demand for ‘Research and development managers’, ‘Actuaries, economists and statisticians’, ‘Electronics engineers’, and ‘Natural and social science professionals n.e.c.’ is greater within GSE regions
  • Demand for the majority of technical and engineering occupations is greater in regions outside of the GSE.
  • Demand for IT and programming roles appears to be highest in London compared to other regions
  • Demand for ‘Higher education teaching professionals’ is slightly higher in regions outside of the GSE on average.

Differences in demand for these occupations are likely related to the industrial makeup across different regions

Table 8 in the accompanying data tables presents a snapshot of the regional distribution of demand for R&D occupations in 2022. It compares their ‘GSE ratio’. The GSE ratio represents the ratio of demand for a given R&D occupation in GSE regions, to that in non-GSE regions.

Salary is an important consideration for many R&D workers. It can act as a strong push or pull factor for where talented individuals decide to work. The majority of job postings in the dataset contain an associated ‘minimum annual salary’ advertised by employers. Table A11 in the accompanying Annex data tables uses this data to present the average salary for R&D occupations, broken down by year and region. Greater London R&D job postings advertise the highest salaries on average. The growth in this average (mean) has been almost £7,000 since 2013. Despite this, postings in the North West have seen the biggest growth in salary for R&D roles since 2013. There has been a £7,000 increase over the 10-year period, closely followed by the North East with a similar increase. Salary growth in Greater London was the largest since 2019. However, so too was salary growth in the GSE on average compared to non-GSE regions. Overall, the gap in salary for R&D roles between the GSE and non-GSE has only narrowed slightly in the past 10 years. This may play a role in the attraction and retention of skilled R&D workers.

Other factors that can also contribute to the salary gap include:

  • the demand for particular skills
  • the level of competition among employers within sectors
  • cost of living considerations

Salaries also vary across R&D occupations. Figure 6 presents the trends in salaries across core R&D occupations compared to broader R&D occupations and non-R&D job postings. The majority of R&D occupations display average salaries notably higher than non-R&D job postings. Technicians are the exception. Their salary, and change in salary, since 2013 is comparable to non-R&D job postings. IT/programming occupations show not only the highest salaries, over £55,000 on average in 2022, but also an apparent spike in salary growth since 2019. This spike may be due to the increased demand for IT/programming skills during the pandemic. This trend is less evident in other R&D occupations. Taking this in combination with the notable increase in demand for IT/programmer roles as a share of demand for all R&D occupations since 2019 (as shown in Table 7 in the accompanying Annex data tables), there may be labour/skills shortages to fill.

Figure 6: Mean salary by occupation group for core R&D occupations, compared to broader R&D occupations and non-R&D job postings (2013 to 2022)

Source: Analysis of Lightcast microdata

Qualitative evidence – relevant findings from interviews with R&D employers and employees

According to respondents, the technician profession has limited visibility and recognition. Furthermore, there is a lack of awareness about technical career paths and the contribution technicians can make to the R&D ecosystem. Recent initiatives, such as the TALENT programme[footnote 84], have started to address these issues. Moreover, specialist technicians are not as well compensated and rewarded in higher education as they are in large corporations in the private sector.

“Technicians are a group who traditionally lacked visibility and recognition in the R&D workforce. So, when you’re thinking about new talent pipeline of technicians, for example, a lot of young people don’t know these careers and roles even exist in our institutions. I think that’s a key challenge. Coupled with that, we’ve had a real lack of sector strategic understanding of what technicians do and how they contribute to the R&D ecosystem, until recent initiatives have begun to address that.” (University partnership representative)

5.4 Education, qualifications and skills

Though we define R&D as a single workforce, the occupations that comprise it require a diverse and often highly specialised set of skills. It is essential that these skill needs are met for the UK to solidify its place as a global science and technology superpower. This section presents the findings from the analysis on skills and qualifications using Lightcast data.

5.5 Qualifications

Lightcast job postings data provides insight into the level of qualifications employers expect through a ‘minimum NQF level’ value. NQF (National Qualifications Framework) refers to the levels assigned to various qualifications. This includes lower-level qualifications, such as A levels (level 3), to honours degrees (level 6), and up to the highest level of qualifications, such as doctorates (level 8). This gives us an idea of the level of education or training that employers expect from applicants for a job vacancy. In this section, we estimate the average NQF level for each R&D occupation group. We also look at trends over time compared to broader R&D occupations and non-R&D postings.

Figure 7 shows that the average (mean) minimum level of qualification expected for R&D job postings is markedly higher than non-R&D postings. The notable exception is technician roles. These job postings have the lowest average minimum NQF level of any group. Technicians have a stable average minimum NQF level of 4 since 2013. This potentially suggests a fixed pathway through apprenticeships (The TALENT Commission, 2022[footnote 85]). However, it should be noted that technicians in the R&D group typically have higher NQF levels than specific occupations within the non-R&D category, which is a very diverse group.

The ‘core’ R&D occupations with the highest average minimum NQF levels in 2022 were:

  • scientists
  • R&D managers
  • business research/analysts

Perhaps unexpectedly, job postings for HE teaching professionals and engineers displayed a lower average minimum qualification level requirement than other ‘core’ R&D occupation groups, excluding technicians. This difference could be due to data limitations and other factors, such as industry standards or different pathways to these roles.

Additionally, we compared average minimum qualification levels in job postings for occupation groups in Figure 7 with average salary data in postings by occupation group in Figure 6. It became clear that a higher average minimum qualification level does not necessarily translate to higher average pay. Despite having the highest average NQF level of any occupation group since 2013, job postings for scientists are second only to technicians in the lowest average salary across occupation groups.

Figure 7: Average qualification level for core R&D occupation groups over time, compared to broader R&D and non-R&D postings (2013 to 2022)

Source: Analysis of Lightcast microdata

5.6 Skill demand

R&D sector level

Figure 8 plots the demand for the 4 different skill types – the highest-level grouping of skills – across core R&D, broadly defined R&D occupations (excluding core occupations), and non-R&D job postings. Specialised skills (excluding software skills) make up the largest share of skill demand for all occupations, particularly in non-R&D job postings. There is a notable difference between skill type demand for core R&D occupations and non-R&D occupations. There is a marked difference in specialised software skills as a proportion of total skill demand. Specialised software skills make up approximately 23% of total skill demand for core R&D occupations, compared to under 5% for non-R&D occupations. Core R&D skill demand is the most specialised of any group, with around 80% of all skills specified in job postings relating to some form of specialised skills (software or otherwise). In contrast, general skills make up a 10 percentage point larger share of non-R&D skill demand. The share of general software skills in overall skill demand is lowest in core R&D occupations. Given that IT/programming occupations make up nearly 50% of core R&D job postings, these figures are to be expected. For broader R&D occupations, demand for each of these skill types appears to be somewhere between the values observed for core R&D and non-R&D occupations. However, overall, they appear closer to core R&D values.

Figure 8: Breakdown of overall skill demand for core R&D, broader R&D occupations, and non-R&D postings by type of skill (highest level grouping) (2022)

Source: Analysis of Lightcast microdata

Job postings from 2022 provide the most up-to-date snapshot of specific skills in demand for core R&D occupations. Table 9 presents the top 25 skills for core R&D postings by their proportion of total skill demand. The table shows that ‘Communication skills’ rank first by a notable margin. Other general skills, including ‘Teamwork’ and ‘Research’, make up the top 5 skills in demand. This aligns with findings from a recent literature review. This review notes a future demand for a blend of analytical/creative, interpersonal, self-management, emotional intelligence, and leadership/management skills, particularly with the emergence of new technologies (Taylor et al., 2022) [footnote 86].

Table 9: The top 25 in-demand skills as percentage of total skill frequency for core R&D occupations (2022)

Skill Type Skill cluster % of demand
Communication Skills General skill General 2.7%
Teamwork / Collaboration General skill General 1.8%
Planning General skill General 1.5%
Problem Solving General skill General 1.4%
Research General skill General 1.3%
Software Development Specialised software skill Software Development Principles 1.2%
Software Engineering Specialised software skill Software Development Principles 1.2%
SQL Specialised software skill SQL Databases and Programming 1.1%
Python Specialised software skill Scripting Languages 1.0%
Corporate Social Responsibility Specialised skill Regulation and Law Compliance 1.0%
Project Management Specialised skill Project Management 1.0%
Writing General skill General 0.9%
Java Specialised software skill Java 0.9%
Microsoft C# Specialised software skill Microsoft Development Tools 0.9%
DevOps Specialised skill Software Development Methodologies 0.9%
Detail-Orientated General skill General 0.9%
JavaScript Specialised software skill JavaScript and jQuery 0.8%
Organisational Skills General skill General 0.8%
Quality Assurance and Control Specialised skill Quality Assurance and Control 0.7%
Creativity General skill General 0.7%
Budgeting Specialised skill Budget Management 0.7%
Microsoft Excel General software skill Microsoft Office and Productivity Tools 0.7%
Teaching Specialised skill Teaching 0.6%
Customer Service Specialised skill Basic Customer Service 0.6%
.NET Specialised software skill Microsoft Development Tools 0.6%

Source: Analysis of Lightcast microdata.

General skills are present across all occupations. A majority (75%) of respondents to the 2022 R&I workforce survey ranked communication as the most important skill[footnote 87]. Though ‘Communication skills’ also rank highly in demand outside of R&D occupations, ‘Research’ and ‘Problem Solving’ are more uniquely sought after in R&D job postings. These skills rank much lower when looking at the top skills for non-R&D occupations using the same method. IT/programming occupations make up almost half of all core R&D job postings. It, therefore, follows that a number of these high demand skills are scripting/coding languages or skills related in some way to software development.

High demand for ‘creativity’ is interesting, as it reflects the widely held belief that creativity is one of the most important skills for R&D. It enables individuals to develop useful products, processes, or services from the ideas/conceptualisation phase (Belt et al., 2021[footnote 88].

Qualitative evidence – relevant findings from interviews with R&D employers and employees

Interview responses reflected the mixture of technical and ‘soft’ skills found in high demand through the job vacancy analysis. Digital, data science, and cyber skills are becoming increasingly important. This is as a result of the emergence of new technologies and the availability of vast amounts of data in an economy undergoing rapid change. Most interviewees highlighted the significance of digital competence regardless of one’s area of expertise:

“The world is becoming increasingly digital, but digital is not the answer to everything. However, we need a workforce that is digitally competent. [It] doesn’t have to be able to do coding, whatever. But it has to be able to understand where the potential for digital and particular things like AI is seen as being the panacea to all ills. But, actually, how to implement that in practice is a huge thing.” (Health technologies representative)

According to a technology transfer engineer, interpersonal skills are also crucial in engineering. These include the ability to:

  • interact with others
  • comprehend their needs
  • inspire and motivate
  • demonstrate leadership
  • exhibit communication skills

The respondent described how spending time as an apprentice in an automotive company gave her the opportunity to learn how to function in a work environment and cooperate with individuals with different personalities. She found these ‘people skills’ (especially communication skills) particularly useful when her role required interaction with other businesses.

In a similar vein, a HEI senior technician (2) identified the following as the most important skills for his role:

  • listening skills
  • communication
  • project management
  • strategic thinking

“First is good listening skills and communication skills. It’s very important. Empathy is very important as well. … You’ve got to be able to understand what support they [people] need and be able to deliver. Project management is another skill, whether it’s fixing a piece of kit to actually building a new lab, you need to be able to project manage. And have a strategic insight or be able to strategize in my role – I think we need to think for the next 5 years, which is what I tend to do.” (HEI senior technician 2)

Finally, an important message emerged from the interviews. Innovation has greater scope than simply developing new products or services. Because of this, creative thinking in the education system should be encouraged from a very early stage of people’s lives.

“Too many people think that innovation is all about design. And then they roll it into “well, we’ve got product development courses”. Innovation is much broader than that. Innovation is about thinking creatively. R&D doesn’t have to be focused around products. Whereas actually, what we need to do is we need to train our young people to be creative, innovative-thinking in whatever subjects that they’re doing. And then that will slowly cascade out into the marketplace. We need to go further back down the chain, and we need to put creativity back into the curriculum at a much, much lower level”. (Manufacturing CEO).

Skills by occupation group

It is important to acknowledge that there is a wide range of occupations within the R&D workforce, and therefore a very broad range of skills used across these occupations. This section investigates the nuance of skill demand across R&D occupations.

To achieve this, specialised skill demand was analysed in each occupation group. Table 10 ranks the specialised skills in highest demand by employers for each of the 7 groups of R&D occupations. Predictably, there is a clear difference in the top skills for each occupation group. The clear top skills for HE teaching professionals are ‘Teaching’ and ‘Lecturer’, at 18.5% and 15.5%, respectively. The top 5 skills for IT/programmers are dominated largely by coding and programming languages, though more nonspecific ‘Software Development’ and ‘Software Engineering’ skills take the top spots. Highly sought-after skills for IT/programmers specifically can vary more than other occupations over time, as programming languages and novel technologies rise and fall in demand. This variation between occupations highlights the diversity of the R&D workforce. It also showcases the unique skillsets desired from R&D workers.

Nevertheless, there are certain skills that show consistently high demand across most R&D occupation groups. The most notable example of this is ‘Project Management’. This ranks as the top in-demand skill for business researchers/analysts, engineers, and R&D managers. It ranks as the second in-demand skill for scientists. The ability to manage projects effectively – coordinating teams, keeping to timelines, and maintaining budgets – appears to be valued across occupations.

Another example of this is ‘Corporate Social Responsibility’ (CSR). CSR appears in high demand for business researchers/analysts, engineers, R&D managers and technicians alike. CSR also grew in demand overall between 2021 and 2022. This increase in demand over time and across occupations suggests a broader global trend. Organisations are increasingly being held accountable for their influence on society and the environment (Ali et al., 2023[footnote 89]).

The high demand for these 2 skills, among more technical skills within each occupation group, demonstrates the diverse range of roles and responsibilities that exist even within occupations in R&D.

Table 10: Top 5 specialised skills per R&D occupation group (2022)

Occupation group Rank Skill % of skill demand
Business researchers/analysts 1 Project Management 1.5%
Business researchers/analysts 2 Corporate Social Responsibility 1.4%
Business researchers/analysts 3 Data Analysis 1.3%
Business researchers/analysts 4 Budgeting 1.2%
Business researchers/analysts 5 Python 1.2%
Engineers 1 Project Management 2.1%
Engineers 2 Mechanical Engineering 1.9%
Engineers 3 Budgeting 1.6%
Engineers 4 Corporate Social Responsibility 1.4%
Engineers 5 Civil Engineering 1.4%
HE teaching professionals 1 Teaching 18.5%
HE teaching professionals 2 Lecturer 15.5%
HE teaching professionals 3 Psychology 1.2%
HE teaching professionals 4 Curriculum Development 1.2%
HE teaching professionals 5 Working With […] Mental Health 1.0%
IT/programmers 1 Software Development 2.4%
IT/programmers 2 Software Engineering 2.4%
IT/programmers 3 SQL 2.3%
IT/programmers 4 Python 1.9%
IT/programmers 5 Java 1.9%
R&D managers 1 Project Management 2.9%
R&D managers 2 Budgeting 2.5%
R&D managers 3 Corporate Social Responsibility 1.6%
R&D managers 4 Market Research 1.5%
R&D managers 5 Quality Management 1.4%
Scientists 1 Chemistry 3.1%
Scientists 2 Project Management 1.6%
Scientists 3 Experiments 1.3%
Scientists 4 Biology 1.2%
Scientists 5 Biochemistry 1.2%
Technicians 1 Cleaning 2.1%
Technicians 2 Corporate Social Responsibility 1.9%
Technicians 3 Quality Assurance and Control 1.9%
Technicians 4 Scheduling 1.7%
Technicians 5 Customer Service 1.5%

Note: The table lists the top 5 specialised skills for each R&D occupation group (based on our ‘core’ definition), ranked by their proportion of total skill demand. Source: Analysis of Lightcast microdata

5.7 Mapping the mismatch: A comparative exploration of skills shortages and gaps in R&D-intensive sectors across UK regions

This section presents an analysis of skills shortages and gaps. It distinguishes between 2 broad sector groups: R&D-intensive sectors versus other sectors (see Section 8.6). We describe the causes and effects of these shortages and gaps. We also explore employers’ mitigation strategies. The empirical basis for this exploration is the Employer Skills Survey (ESS) 2019. This was the latest data available at the time of writing this report.

Establishments and employment: regional, size, occupational, and sector analysis

The Employer Skills Survey provides insights into establishments and employment in R&D compared to non-R&D sectors. In this survey, an establishment refers to a branch or site of an organisation with at least 2 staff on payroll. This analysis finds that the main R&D industry divisions as a percentage of total R&D establishments are:

  • ‘computer programming, consultancy & related activities’ comprising 62.5% of all R&D establishments
  • ‘manufacture of machinery and equipment n.e.c.’ at 10.0% of total R&D establishments
  • ‘manufacture of computer, electronic and optical products’ at 5.6%

See:

for more details on the methodology applied.

The estimated total number of R&D related establishments was 80,100 in 2019. This represents 4.4% of all 1.83 million establishments in England, Wales, and Northern Ireland. The distribution of these R&D establishments is not the same across the regions. The South East holds the largest proportion (20.7%) of all R&D establishments. This region is followed closely by London (17.8%) and the East of England (12.4%). The combined figures for these 3 regions account for just over half of the total number of establishments. This aligns with the fact that these regions absorbed 54% of the total gross domestic R&D expenditure in the UK in 2019 (ONS, 2021b[footnote 90]). This picture indicates a considerable concentration of firms conducting R&D activities in these regions. This is potentially due to more intensive industry presence and investment. In contrast, the proportion of R&D establishments is markedly lower in Wales and Northern Ireland. In 2019, these 2 nations were responsible for just 4.2% of the total expenditure on R&D in the UK. Table 11 in the accompanying data tables provides a detailed breakdown of establishments by region and sector group (‘R&D sectors’ versus ‘other sectors’), using weighted figures.

As well as the distribution of R&D establishments, there is also regional variation in the size of R&D establishments in terms of their employment share. Overall, R&D establishments typically employ larger numbers of staff. Specifically, 6.7% of all R&D establishments have a workforce exceeding 50 employees. This is in contrast to just 4.7% in other sectors (see Table A12 in the accompanying Annex data tables). This percentage is highest in the North East, where 10.5% of R&D establishments have a workforce exceeding 50 employees, and lower in London (5.1%), East of England (5.4%), and the South East (5.8%). This could be due to differences in the nature of the R&D sectors in these regions. For example, London is less suited to heavy industry and large-scale manufacturing due to its urban nature.

Figure 9 illustrates regional variations in the establishment and employment shares in R&D sectors. For instance, R&D businesses and organisations in the South East account for 5.8% of total establishments and 7.3% of total employment in the region. Meanwhile, in the West Midlands, the shares are lower at 4.2% and 5.9%, respectively. The most substantial gaps between the employment and establishment shares in the R&D sectors are observed in the North East. As a percentage of regional totals, there is a gap of 3.4 percentage points. A larger gap, where the employment share exceeds the establishment share, suggests the presence of larger firms with higher employment numbers. In other words, these regions have fewer but larger R&D establishments relative to other regions. This indicates a concentration of employment within these larger organisations. Conversely, a smaller gap or even a reversed gap (as in the case of London) suggests a prevalence of smaller firms in R&D sectors. Again, these figures highlight regional disparities in the scale of R&D establishments, possibly influenced by factors, such as:

  • industrial structure and focus
  • regional policies
  • infrastructure
  • local labour market conditions

Figure 9. Establishment and employment shares (%) in R&D sectors across UK regions

Note: The figures in the graph are weighted based on the number of establishments and employees. Source: Employer Skills Survey 2019, authors’ own calculations

5.8 Vacancies, skills shortages and skills gaps

The Employer Skills Survey offers a detailed insight into job vacancies across various sectors. This allows for a comparison between R&D establishments and other types of businesses regarding:

  • hard-to-fill positions
  • vacancies arising from skills shortages
  • areas where skills gaps exist

The objective of this analysis is to understand how these challenges differ based on the sector and geographical region of the establishments. As mentioned earlier, this analysis uses data from the 2019 collection, which was the most recent data available at the time of analysis. The 2022 survey results have since been published. The 2019 figures are not directly comparable to the 2017 or 2022 figures due to different geographic coverage (with Scotland not included in the 2019 survey). The long-term trend indicates an increase in vacancies over this period. See Section 8.6 for methodology details.

Vacancies and hard-to-fill vacancies: regional and sector analysis

The total number of vacancies for all sectors in England, Wales and Northern Ireland was 877,000 in 2019, with 5.4% of these vacancies relating to roles in R&D establishments. This varied by region. The South East and London accounted for the largest proportion of R&D sector vacancies (25.3% and 19.2% of total vacancies were in these regions respectively). Northern Ireland and Wales accounted for the lowest (2.2% and 2.6% respectively). Understanding the distribution of hard-to-fill vacancies is important for informing strategies to address recruitment difficulties. R&D sectors accounted for 5.4% of overall hard-to-fill vacancies in the UK in 2019. The highest proportion of these was in the South East (24.0% of hard to fill vacancies in R&D sectors were in the South East). The East Midlands also showed a proportion of hard-to-fill vacancies that far outstripped its share of total vacancies (11.7% versus 6.0%). This could suggest more specialised roles that are more challenging to fill due to skill requirements.[footnote 91]

Skills-shortage vacancies and skills gaps: regional and sector analysis

Of the total 214,300 skills-shortage vacancies in 2019, the R&D sectors accounted for 15,600 (7.3%). Overall, skills-shortage vacancies represented 68% of all hard-to-fill vacancies. This share is higher among R&D intensive sectors (92%), revealing their increased difficulties in finding people with the right skills, qualifications, and experience. In R&D sectors, the South East region emerges as a focal point again, making up 23.9% of all skills-shortage vacancies. Notably, the East Midlands sees a disproportionate share of skills-shortage vacancies in R&D sectors (12.5% of R&D skills-shortage vacancies in the UK occur in this region) relative to its total share of these vacancies across all sectors (6.7%). This suggests there is a high demand for specific R&D-related skills. It also suggests employers face increased challenges to meet their skills needs. In Wales and Northern Ireland, skills-shortage vacancies in the R&D sectors comprise only 1.8% of the total number of R&D skills shortage-vacancies. This mirrors a smaller skills shortage issue in these nations and a smaller R&D sector overall. Table 12 in the accompanying tables provides further detail of the spread of skills-shortage vacancies across the UK (excluding Scotland).

5.4% of employees with skills gaps are found in the R&D sectors. The East of England and South East are prominent contributors to skills gaps in R&D sectors (17.0% and 17.8% of workers in R&D sectors with skills gaps, respectively, are in these regions). The West Midlands displays a disproportionately high proportion of skills gaps in R&D sectors compared to its proportion across all sectors (7.4% overall compared to 14.1%). This signifies a mismatch between the skills of the current workforce and the employers’ skills demands. London, in contrast, holds a 15.3% share of skill gaps across sectors but only 3.1% in R&D sectors (see Table 13 in the accompanying data tables for further details). This may be due to the attractiveness of London for highly skilled workers, allowing establishments to meet skills needs.

5.9 Incidence and density of vacancies, hard-to-fill vacancies, skills shortages, and skills gaps

The R&D sectors generally experience a higher rate of vacancies. This suggests an elevated demand for talent in this field. For instance, in the North East and the South East, the rate of vacancies in the R&D industries is 20.8% and 20.6%, respectively, compared to 14.6% and 18.3% in other sectors.

When considering hard-to-fill and skills-shortage vacancies, we see a similar trend. A particularly telling case is the South West. Here, the R&D sector reports a 11.5% incidence of hard-to-fill vacancies and a 10.3% incidence of skills-shortage vacancies. Nevertheless, the incidence of skills gaps, which reflect employees’ lack of skill within their role, presents a different dynamic. In the North West, the East Midlands, and Northern Ireland more employers experience skills gaps in the R&D sectors than in other sectors. Other regions, such as London, buck this trend. The capital experiences a low 4.4% skills gap incidence for the R&D sectors. In contrast, Northern Ireland’s highest skills gap rate (17.9%) likely signals the need for further investment in workforce development and training within its R&D sector. Table A15 in the accompanying Annex distinguishes between R&D-intensive sectors and other sectors to provide a detailed overview of the rate of:

  • vacancies
  • hard-to-fill vacancies
  • skills-shortage vacancies
  • skills gaps across regions

Overall, the densities of vacancies, hard-to-fill vacancies, and skills gaps do not differ significantly between the R&D-intensive and other sectors. In contrast, a marked variation appears in the density of skills-shortage vacancies, where the R&D sectors surpass other sectors (32.9% versus 24.0%). This finding is important as it emphasises the distinct difficulties faced by R&D employers in filling vacancies with adequately skilled workers. As a result, this pattern highlights the ongoing challenge for the R&D sectors to locate and secure skilled talent (both from within the country and from abroad). This suggests a suitable call for strategies and policies to address this issue.

While the density of vacancies appears stable across regions, the differences become clearer in terms of hard-to-fill and skills-shortage vacancies. In particular, in the East Midlands’ R&D sectors, an alarming 64.7% of vacancies are hard-to-fill, and 63.4% are skills-shortage vacancies. This reflects heightened recruitment challenges. This issue of skills-shortage vacancies may affect a significant number of businesses in the East Midlands, including those leading in their sectors.

Skills gaps show similar regional variations. Interestingly, London’s R&D sectors demonstrate a strikingly low density of skills gaps (1.1%). This may be due to:

  • better access to training resources
  • higher talent proficiency levels
  • a large pool of highly skilled labour to draw from

On the other hand, the East of England, the North West, and the West Midlands have higher densities of skills gaps (7.8%, 6.5% and 6.3%, respectively). This possibly points to deficiencies in training and development opportunities. Table 14 in the accompanying tables provides detailed information about the density of:

  • vacancies
  • hard-to-fill vacancies
  • skills-shortage vacancies
  • skills gaps in R&D and other sectors across regions

The density of skills shortage vacancies and skills gaps also varies with establishment size. Smaller R&D establishments (with 5 to 9 or 10 to 24 number of employees) witness higher densities of skills-shortage vacancies (41.5% and 41.2%, respectively) than larger ones.

Looking at skills gaps, establishments with 25 to 49 and 100 to 249 employees reported higher densities of skills gaps in the R&D sectors. This implies a likely need for skills improvement initiatives aimed at medium-sized businesses (see Table A16 in the accompanying Annex tables).

Figure 10 reveals regional differences in the proportion of R&D sectors’ vacancies and skills gaps relative to regional totals (that is, as a fraction of the respective regional totals across both R&D and non-R&D sectors). For instance, the North East and East Midlands regions experience greater recruitment challenges. This is reflected in their above-average shares of hard-to-fill and skills-shortage vacancies. Meanwhile, the West Midlands and the East of England show high proportions of skills gaps, implying substantial development requirements in their R&D workforce. Conversely, London demonstrates a very low percentage of skills gaps, suggesting the presence of a more effectively trained R&D workforce. This is in line with the findings discussed earlier. These differences highlight the diversity of skills-related challenges across regions. It also underlines the importance of tailored, region-specific workforce development strategies in R&D-intensive sectors.

Figure 10. Shares (%) of vacancies, hard-to-fill-vacancies, skills-shortage vacancies, and skills gaps in R&D-intensive sectors across UK regions

Note: The figures in the graph are weighted based on the number of employees. Source: Employer Skills Survey 2019, authors’ own calculations

Qualitative evidence – relevant findings from interviews with R&D employers and employees

A health technologies representative was interviewed about regional differences in establishments’ ability to fill vacancies and skills needs. They mentioned that some companies struggle to recruit even when offering high salaries. There are perceptions that the sector is small in the West Midlands and provides few opportunities for advancement. One reason for this is that the skills levels within many companies in the sector are not especially high. This discourages individuals from relocating to the region.

“People have no interest in coming here. They offer more and more money, and they still can’t fill the posts. So, it’s a massive problem for us. I think the options really are to grow your own [talent], rather than to try and attract them from elsewhere. Because trying to bring people who have the experience of working in Cambridge, where, you know, you walk down the street, and you fall over 5 more people like yourself, that just doesn’t happen here. And that is a big barrier for us.” (Health technologies representative)

R&D employers were interviewed about the higher rate of skills shortage vacancies in smaller R&D establishments. They were also asked about higher rates of skills gaps in medium-sized establishments highlighted in this analysis. Their responses suggested that skills needs of SMEs may be considered to a lesser degree by universities.

“I don’t believe that universities fully understand some of the needs of SMEs and they continue to churn out very, very good engineers. But how relevant are those engineering skills to SMEs? And will those engineers actually go and work for SMEs? And maybe that’s where colleges are better? Because colleges (certainly the local colleges here, Dudley College, Walsall College) are probably working much, much closer with businesses to say ’what do you actually need?’” (Manufacturing CEO)

The interviews with R&D workers also showed that SMEs are less likely to send their employees to training programmes. This is because of the time and money required to participate in such programmes. It is also difficult to measure the direct and immediate benefits of training. This may be linked to the higher rates of skills gaps identified in medium-sized businesses.

5.10 Skills difficult to obtain from applicants

Figure 11 illustrates the skills that employers find difficult to obtain from job applicants at the country level (excluding Scotland). It is evident that there are differences in how common certain skills between the 2 sector groups are. R&D-intensive industries generally exhibit higher demand for specific skill sets.

The most important skill lacking in applicants to R&D-intensive sectors is specialist skills or knowledge. In 2019, 77% of establishments with skills shortage vacancies in R&D sectors revealed this deficit. This illustrates the complex and specialised nature of jobs in the R&D sectors and the importance of industry-specific knowledge. Furthermore, the ability to solve complex problems is reported as a hard-to-fill skill by 57% of R&D establishments. This reflects the critical need for problem-solving abilities to address complex research and development issues. Similarly, 50% of establishments in R&D-intensive sectors with skills shortage vacancies report a shortage in applicants with advanced or specialist IT skills. This emphasises the digitally-driven nature of the sector. Looking into specific IT skills, the most frequently reported as difficult to obtain by R&D establishments compared to non-R&D (see Figure A7 in Section 9) include:

  • specialist software or hardware/internal systems skills
  • application programming and development skills
  • building/maintaining IT systems and networks
  • data analysis/analytics/data science

In contrast, there is a lower demand in R&D sectors for skills such as ‘knowledge of how the organisation works’ and ‘writing instructions, reports, etc.’. This suggests a greater emphasis on technical abilities and specific knowledge, rather than organisational familiarity and general writing skills. It is noteworthy that some basic skills, such as computer literacy and basic numerical skills, are perceived as less problematic in R&D sectors compared to other sectors. This indicates a generally higher baseline of these skills among applicants in the R&D industry.

Figure 11. Prevalence of technical/practical skills difficult to obtain from applicants by sector group

Note: The graph shows the weighted proportion of establishments reporting a particular skill shortage. Base: Establishments with skills-shortage vacancies. Figures refer to England, Wales, and N. Ireland (excluding Scotland). Source: Employer Skills Survey 2019, authors’ own calculations

Figure 12 shows the prevalence of soft skills found challenging to acquire from job applicants. The data highlights some important areas that are particularly relevant to R&D sectors. The top on the list is the ability to manage one’s own time and prioritise tasks. This was reported as a shortage by 47% of R&D establishments with skills shortage vacancies in 2019. This illustrates the importance of effective self-management and organisation, possibly due to the project-based nature of many R&D occupations. Customer handling and team working skills are also identified as weaknesses, identified by 35% and 34% of R&D establishments, respectively. It should be noted that these skills are also valued highly by employers in non-R&D sectors. Lastly, ‘don’t know’ responses are higher in R&D-intensive sectors than in other sectors (10% versus 6%). This reflects a likely uncertainty in identifying soft skills gaps within these establishments. This highlights the increased need for a better understanding and communication of the soft skills requirements in R&D sectors.

Figure 12. Prevalence of soft/people skills difficult to obtain from applicants by sector group

Note: The graph shows the weighted proportion of establishments reporting a particular skill shortage. Base: Establishments with skills-shortage vacancies. Figures refer to England, Wales, and N. Ireland (excluding Scotland). Source: Employer Skills Survey 2019, authors’ own calculations

5.1.1 Consequence of hard-to-fill vacancies on employers and actions taken to address these

Difficulty in filling vacancies can result in negative consequences for employers. Figure A8 in Section 9 shows the most frequently reported consequences for employers in R&D sectors were:

  • increased workload for other staff (reported by 88% of R&D establishments with hard-to-fill vacancies)
  • delay in developing new products or services (56%)
  • difficulties meeting customer service objectives (51%)
  • increased operating costs (45%)

Delay in developing new products or services at 56% is markedly higher than the figure for non-R&D establishments of 34%. This suggests a particular disruption for R&D establishments considering the central role of innovation and product development in the R&D sectors.

Establishments reported employing a range of measures to address hard-to-fill vacancies. This is shown in Figure A9 of Section 9. In R&D-intensive sectors, 42% of establishments reported adopting new recruitment methods or channels. This percentage is higher than the 35% observed in other sectors. Other sectors (37%) are more likely than R&D-intensive sectors (29%) to increase advertising or recruitment expenditures. This could imply that other sectors have more targeted recruitment strategies and marketing campaigns in place allowing them to compete effectively with R&D sectors for highly skilled talent. Notably, a greater proportion of R&D establishments (13%) have increased or expanded their trainee programmes than other sectors (7%) to overcome hard-to-fill vacancies. This strategy may be indicative of a long-term plan to cultivate required skills in-house. This is a logical step given the requirements for specialised expertise in R&D. Fewer R&D establishments (8%) reported doing nothing in response to these hard-to-fill vacancies compared to other sectors (12%). This suggests a relatively more proactive approach in the R&D sectors to address talent shortages.

Qualitative evidence – relevant findings from interviews with R&D employers and employees

Employers described specific strategies that can effectively improve equal opportunities in the R&D workforce. This could, in turn, contribute to reducing how common hard-to-fill vacancies in R&D sectors are. Specifically, a representative of an energy infrastructure organisation claimed that adjusting the language of job advertisements, creating more “jobs of purpose”, and facilitating flexible working arrangements would likely attract more women to engineering and broader R&D-related professions.

“One of the things we’ve really noticed, and the research shows this, is women are more driven by jobs of purpose than men are. The way you word adverts can inadvertently be more appealing or less appealing to different diversities, like ethnic minorities or genders. So, we’ve really focused a lot on the language that we put into our job adverts. Particularly, rather than just listing the requirements, we focus on the purpose using words like ‘design’ and ‘create’. Some of that much more creative language is much more appealing to females than it is to males. Flexibility and flexible working arrangements are also much more important to females.” (Energy infrastructure organisation representative)

Some employers encourage inclusive recruitment through anonymous applications and favouring women when candidates are equally qualified to increase diversity in their organisation.

“When it gets to recruitment, you need every employer to be truly inclusive. So, when we do our graduate recruitment, we take your name off, your picture off, take your school education qualification [off]. It’s completely person neutral, and we just focus on your personal statement.” (Large engineering company director)

“It is a heavily male dominated system and what we are doing is not a secret – when we interview people for the internship or other things, if the 2 candidates are very close to each other, we go for the female one.” (Digital engineering director)

Investing in recruitment and marketing strategies can help technology, engineering and aerospace companies attract STEM graduates. It can also mitigate the effects of competition from other industries (for example, the financial services sector). However, good salaries and an established brand name are only part of the story. Firms can attract and retain R&D talent if they preserve and improve the value they contribute to society.

“I don’t think all STEM graduates are motivated by salary. … I suspect there are a few STEM graduates who choose those degrees because they know they’re appealing to the big-paying financial services industry, and probably always had that intention.… But I think others are really motivated now by making a difference, you know, sustainability. So, I think it can help to look at the salary package and the brand, but also to really do a lot to show what these careers offer – the value that you can add to society is quite appealing as well.” (STEM institute director)

5.12 The internal skills challenge: skills gaps and their causes, implications, and corrective responses

In both R&D and other sectors, skills gaps are frequently attributable to temporary conditions. These conditions are expected to lessen as employees’ tenure in an organisation increases. Specifically, partial completion of training (68% and 63%, respectively) and newness to the role (63% each) were identified as the main causes of skills gaps in the 2 sector groups (see Figure 13). However, there are pronounced differences in some categories. For instance, R&D-intensive sectors appear more affected by skills gaps caused by the development of new products and services (26%) and the introduction of new technology (24%). These are both higher than in other sectors (17% and 20%, respectively). The response ‘staff lack motivation’ contributes less to skills gaps in R&D-intensive establishments (23%) than in other sectors (30%). This likely indicates a more engaged R&D workforce. Furthermore, a higher proportion of R&D employers attribute skills gaps to a lack of appropriate training (32%) than their non-R&D counterparts (25%). This implies a need for tailored training strategies in R&D sectors.

R&D sectors and other sectors vary in the types of skills where gaps exist. For technical/practical skills, for example, skills gaps in advanced or specialist IT skills were reported at a far higher rate by R&D establishments (42%) than other sectors (21%). Looking at IT skills specifically, R&D establishments reported higher skills gaps than other sectors in:

  • advanced Microsoft Office skills (23% compared to 18%)
  • application programming and development skills (17% compared to 5%)
  • building/maintaining IT systems and networks (12% compared to 2%)

Conversely, other sectors reported higher skills gaps than R&D establishments in:

  • more basic IT skills, such as basic Microsoft Office skills (33% compared to 18%)
  • foundation digital skills (20% compared to 9%)

This signals a broader requirement for fundamental digital competency in non-R&D sectors.

The most common soft skills requiring improvement among staff in both R&D and other industries are:

  • time management
  • task prioritisation
  • teamwork

Customer handling skills are more common in ‘other sectors’ (47%) than they are in R&D sectors (34%). This highlights the varied interpersonal skills requirements across industries. For more details, see Figures A10 to A12 in section 9.

Figure 13. Main causes of skills gaps among staff by sector group

Note: The graph shows the weighted proportion of establishments reporting a particular cause of skills gaps. Base: Establishments with skills gaps. Figures refer to England, Wales, and N. Ireland (excluding Scotland). Source: Employer Skills Survey 2019, authors’ own calculations

Figure 14 demonstrates that the most significant consequence of skills gaps is increased workload for other staff. More than half of establishments in both sector groups report this. R&D sectors witness a more pronounced delay in developing new products or services (28%) compared to other sectors (15%). This is in part due to their fundamental reliance on advanced, specialised skills.

Figure 14. Implications of skills gaps (prompted) for employers by sector group

Note: The graph shows the weighted proportion of establishments reporting a particular implication. Base: Establishments with skills gaps. Figures refer to England, Wales, and N. Ireland (excluding Scotland). Source: Employer Skills Survey 2019, authors’ own calculations

Figure A13 in Section 9 shows that employers report applying a range of countermeasures to fix skills gaps and their consequences. Most commonly, both R&D and other sectors resort to enhanced training activities. R&D establishments show a higher tendency for this (69% for R&D establishments with skills gaps compared to 64% for others). This signifies a proactive, internal adjustment approach in R&D establishments. R&D establishments were also more likely than other sectors to hire non-UK nationals (11% compared to 8%). As was the case with hard-to-fill vacancies, they are less likely to report doing nothing (14% compared to 18%).

Qualitative evidence – relevant findings from interviews with R&D employers and employees

There was a consensus among interviewees that investing in training and upskilling/reskilling the current workforce will remain critical for meeting future skill requirements. The significance of management and leadership skills is often overlooked. This has implications for the training of the current and future workforce.

“I think there’s a lot to be done around the upskilling of the current workforce. I think we often don’t talk about that. One of the most important bits that is often overlooked is the leadership and management skills – we don’t seem to see that as an academic trend and a need to keep on that lifelong learning. And unless you’ve got the right leadership at the top, then all this other stuff is just not going to flow.” (Representative of a manufacturing organisation)

Several interviewees identified a role for universities in developing short/continuing professional development (CPD) courses for upskilling. However, the general consensus was that, to date, universities have not fully grasped this role. The issue of a disconnect in timescales between universities and businesses, and concerns about market demand are also apparent in relation to the development of short/ CPD courses. As a health technologies representative noted:

“I’ve had many discussions with universities about continuing professional development needs of the existing workforce, and universities always sympathetic, but don’t, I think, believe there’s a market for them there.” (Health technologies representative)

This same interviewee felt that if universities became more involved in providing CPD, this would have positive outcomes for working with businesses more generally and addressing R&D skills needs:

“And I think if we did get into the mindset that there is such a market for refreshing people’s skills, adding to their skills, and giving them opportunities to learn more, and do more, I think it would change the way of thinking in universities much more radically, in terms of [the] commercial approach to supporting businesses in meeting the skills needs that they will have in the future.” (Health technologies representative)

6. Projections for R&D employment prospects

This section estimates the future of the UK’s R&D workforce. It uses data from the Skills Imperative 2035 research programme[footnote 92]. The Nuffield Foundation funded this programme. We explore forecasted trends in occupations essential for R&D. 

As outlined in Section 8.8, the Skills Imperative data used a range of macroeconomic assumptions and past trends to forecast the size and makeup of the UK workforce between 2021 and 2035 at the 4-digit SOC level. We applied the projected annual jobs growth rates forecast in the Skills Imperative 2035 research to the July 2023 ONS Annual Population Survey employment estimates. We then calculated updated projection figures for employment in R&D occupations.

This analysis includes projections based on 3 scenarios:

  • baseline scenario,
  • technological opportunities scenario
  • human-centric scenario.

The growth in the number of people working in R&D from 2021 to 2023 has so far aligned more closely with the jobs trajectories forecast in the human centric and technological opportunities scenarios than the baseline scenario. However, this may be related to uncertainties in forecasting the post-COVID-19 recovery. Longer term trajectories may vary.

Data limitations mean that individual occupations have high levels of uncertainty in forecasting. They also have broader challenges in making future predictions.

For R&D occupations, there are many factors with potential to influence the trajectory of employment numbers. For example, these include:

  • wider economic growth
  • university sector growth
  • public R&D budgets
  • immigration to the UK
  • the influences of new technologies such as artificial intelligence.

While the number of people employed in R&D is likely to grow, this cannot be forecast with absolute certainty. Uncertainty grows when attempting to forecast the magnitude of any change.

Analysis in this section should be considered indicative of possible futures of the R&D workforce, rather than a precise estimate of what the total workforce size will be. 

6.1 Baseline scenario

The Skills Imperative 2035 ‘Baseline scenario’ used the existing trajectory of technological advancement, political landscapes, demographic changes, and environmental developments at that time to inform future projections. This scenario served as the primary foundation for the projected employment levels between 2021 and 2035. However, it did not predict potential future policy changes that could influence new trends (see Section 8.8 for a description of the Baseline scenario).

Using projections calculated in the Skills Imperative under this baseline scenario, the R&D workforce in the UK is expected to continue to increase between 2024 and 2035. This increase is expected both in terms of absolute numbers and as a proportion of total employment.

According to the Annual Population Survey, an estimated 2.8 million people were employed in occupations essential for R&D in 2023 (see Figure 1a). This represents 8.4% of the total 33.3 million people in employment that year[footnote 93]. By applying the Skills Imperative ‘Baseline’ scenario annual growth rates to the June 2023 Annual Population Survey figures, we can estimate that the number of people in R&D occupations could grow to 3.1 million people in 2035. This would account for 9.1% of total employment and an increase of 372,500 people.

This trend reflects the anticipated increase in demand for R&D workers, such as:

  • programmers and other IT professionals
  • engineers
  • scientists
  • researchers
  • technicians
  • R&D managers
  • higher education teaching professionals

The rising share of R&D employment in total employment indicates that the R&D workforce is expected to expand at a faster rate than the overall labour force. Several factors could encourage this shift, including:

  • the rapid advancement of technology
  • increased emphasis on innovation across all sectors
  • ongoing governmental support for R&D activities

6.2 Technological opportunities and Human-centric scenarios

The ‘Technological opportunities’ scenario recognises the benefits that arise from investing in technologies. These benefits enhance human productivity, smooth the transition to a carbon-neutral economy, and elevate the quality of social services. This scenario projects a higher rate of growth of employment in R&D occupations between 2023 and 2035 than the Baseline scenario. Under this scenario, the total number of people employed in R&D occupations is projected to grow to 3.5 million people in 2035. This would make up 10.1% of UK employment and an increase of 709,200 people.

The ‘Human-centric’ scenario places more emphasis on social services, healthcare, education, and skills less likely to be automated (for example, soft skills). It also maintains some focus on technological and environmental advancements. Under this scenario, the number of people in R&D occupations is also projected to grow to 3.5 million people. This would make up 10.2% of UK employment and an increase of 726,700 people.

Figure 15: Projected number of people working in occupations essential to R&D in the UK from 2023 to 2035 under the Skills Imperative Baseline, Technological opportunities and Human-centric scenarios.

Note: 2023 represents the number of R&D workers as of July 2023 using APS data. 2024 to 2035 represent projection using the growth rates forecast in the Skills Imperative 2021 to 2035 analysis. Source: Annual Population Survey and the Skills Imperative 2021 to 2035, authors’ calculations.

Occupation level growth rates:

The Skills Imperative projections assume the same rate of growth for all occupations within each 2-digit SOC code. By looking at the 2-digit level projections, we can see how different groups of R&D occupations may grow at different rates in each scenario.

Under the baseline scenario, business research professionals, quality control and planning engineers, actuaries, economists, and statisticians are projected to have the fastest rate of growth among R&D occupations. However, under the technological opportunities and the human-centric scenarios, technician occupation groups are projected to have the highest growth rates.

Higher education teaching professionals and traditional R&D occupations, such as scientists, engineers, and IT professionals, are projected to have higher rates of growth in the human-centric scenario than the technological opportunities or baseline scenarios.

Figure 16: Projected percentage increase of R&D occupations in the UK from 2023 to 2035 by occupation group under the Skills Imperative baseline, technological opportunities, and human-centric scenarios.

Source: Skills Imperative 2021 to 2035, authors’ calculations.

Qualitative evidence – relevant findings from interviews with R&D employers and employees

Interviewees were asked about encouraging people into R&D occupations to meet future demand. Most felt that initiatives focusing on post-16 students are often too late. Many individuals will have decided what they want to study by the time that they reach this age.

“The key thing that we found from our research and the work that we’ve done is aspiration. So, if we look at young people, I’m talking about nursery level here, onwards. It’s very much from their first kind of early days in the education system. That, we think is really crucial in terms of aspiration.” (STEM education representative)

“Even before we get to universities, you’ve got to go 3 steps back and start off at school.” (Energy company representative)

The undervaluing (in terms of visibility and recognition) of important R&D professions (for example, technicians and engineers) is a potential barrier to encouraging people into the professions required to meet demand for R&D employment by 2035.

“We in the UK don’t regard engineering as important as probably accountancy or other skill sets, which I think is wrong. If you don’t have good engineers, you don’t have good product, you don’t have a business, you don’t have an economy. So therefore, I think there is a need in the UK to uprate the standing of engineers. […] In Europe, for example, they have a special title, so it’s almost like doctor, to show that they are a professional standard. And they are superbly well regarded in Europe, it’s not the same here. So, we need to do something about that. (Representative of a regional high-tech organisation)

Furthermore, according to interviewees, the relatively lower salaries in STEM positions result in a significant loss of STEM graduates to higher-paying industries, such as banking and finance. This continuing issue with STEM graduates working in less relevant sectors increases the pressure on STEM employers to find people with the necessary skills.

7. Conclusions

The Department for Science, Innovation and Technology (DSIT) aims to accelerate innovation, investment and productivity through world-class science, ensure that new and existing technologies are safely developed and deployed across the UK and drive forward a modern digital government for the benefit of its citizens.

The government’s missions to break down barriers to opportunity and to kickstart economic growth can be enabled through its ambitions on R&D. This includes aims to:

  • strengthen the R&D talent pool
  • ensure inclusivity, diversity, and sustainability of R&D careers
  • attract, support, and retain R&D talent.

This report aims to provide a better understanding of the R&D workforce to help policy making and further support interventions.

Using a range of data sources and methodologies, this report sets out ongoing labour and skills shortages. It outlines the personal characteristics and occupational makeup of the R&D workforce. The report consists of a number of different analytical components (both qualitative and quantitative) to adequately address the different aims of the overall project. This allowed an investigation into the R&D workforce from several angles, from ‘hard-to-fill vacancies’ and their influences on R&D firms, to perspectives on the influence of equality, diversity and inclusion (EDI) issues.

The report outlines the overall likelihood of workers conducting R&D activities and the trends in the activities and profiles of the R&D workforce. The analysis finds that whilst COVID-19 negatively influenced several occupations, the pandemic resulted in an increase in demand for R&D workers. Highly skilled jobs were less negatively influenced, with a growth in the R&D workforce between 2012 and 2023.

The analysis also confirms that there is a strong regional component to the workforce. The highest share of R&D workers in the total workforce is found in the South East (this is averaged over 2012 to 2021). However, the most significant expansion in the R&D workforce can be seen in London. This reinforces London’s status as a hub for R&D). The East of England, Yorkshire and the Humber, and the North East saw the slowest growth in the share of R&D workers. The R&D workforce has been projected to continue growing until at least 2035. This reflects the increase in demand and the rapid advancements in technology in the UK.

When looking at establishments in R&D intensive sectors, similar regional patterns can be seen, with the South East holding the largest proportion of R&D establishments, followed closely by London. The East of England also holds one of the largest proportions of R&D establishments. Together in 2021, these 3 regions (constituting the Greater South East) absorbed 55% of total gross domestic R&D expenditure in 2022[footnote 94]. On the other hand, the analysis finds that the proportion of R&D establishments is lowest in Wales and Northern Ireland. The largest R&D establishments are concentrated in the North East, with smaller R&D establishments found in London, East of England, and the South East. This suggests that the regions with the highest proportion of establishments and workers are largely made up of small to medium enterprises.

The report also describes the average wages of the R&D workforce. On average, R&D workers earn more than their non-R&D counterparts. The hourly wage, however, has only shown small fluctuations in real terms from 2012 to 2021. This is despite there being an increase in the relative supply of highly skilled workers. R&D workers have seen a small wage decrease when accounting for inflation. IT/Programming occupations have shown the highest salaries in R&D since 2019, and technicians the lowest. There are also regional differences in salaries that have only slightly narrowed in the last ten years. Greater London advertises the highest salaries for R&D workers, and the North West has seen the largest growth in salaries. It is important to continually review average wages, as these can work as push or pull factors for people entering or remaining in the R&D workforce. They can also help inform policies and strategies on where there may be disparities.    

7.2 R&D workforce characteristics

The report also outlines the personal characteristics of the R&D workforce. Encouraging a diverse range of people into R&D careers may be important to meeting the UK’s growing demand for R&D workers. In particular, gender disparities emerged with the majority of the R&D workforce being men. Moreover, a gender pay gap exists for R&D workers. Women on average earn less than men. The difference ranges from 9.8% in 2018 to 19.5% in 2016. This gap has also not narrowed between 2012 and 2021. It is, however, lower on average than the pay gap for the overall workforce.

The analysis also highlights that employment in occupations related to R&D varies by ethnic group. More specifically, analysis using the Annual Population Survey (APS) data revealed that Chinese and Indian individuals are more likely than others to be employed in R&D roles. Individuals from Bangladeshi, Black, and Pakistani ethnic groups are less likely to be part of the R&D workforce.

The data also shows that people with long-term health conditions or disabilities are less likely to be employed in occupations related to R&D.

To increase equality, diversity and inclusion in the R&D workforce, this report highlights the need to have better systems to:

  • promote role models
  • raise awareness of STEM careers to address the lack of diversity
  • encourage participation in R&D, for example, through STEM ambassadors

Any potential interventions to increase awareness and exposure to STEM opportunities should be conducted at an earlier stage (for example, at school) to have a greater influence on individuals. This suggests there could be value in implementing further strategies and interventions throughout education that would encourage a diverse group of people to join the R&D workforce. Existing research reveals the positive influence EDI can have on economic growth[footnote 95], [footnote 96], productivity[footnote 97], [footnote 98], [footnote 99], innovation [footnote 100], [footnote 101], [footnote 102], and entrepreneurship [footnote 103].

7.3 Skills demand, shortages and gaps in R&D

This report also explores the changing skills demand within the R&D workforce and the effect this can have on businesses, utilising job vacancy data and the Employer Skills Survey. For each occupation group within R&D, there are clear differences in the top skills sought. This highlights the diversity of work undertaken by the R&D workforce. Notably, R&D skills demand was more specialised than non-R&D occupations, with software skills in particularly high demand. Perhaps unsurprisingly, research and problem solving were more sought after in R&D job postings than non-R&D postings. Project management and Corporate Social Responsibility (CSR) were also consistently sought after. The demand for CSR skills may be in response to the growing global trend to hold companies accountable for their influence on society and the environment. 

The skills gaps reported in the R&D workforce are a useful indicator of the mismatch between the skills of the current workforce and employers’ skills demands. Analysis of the 2019 Employer Skills Survey reveals that many R&D employers have difficulty finding people with the right skills, qualifications, and experience. On average, the minimum level of qualification expected for R&D workers is noticeably higher than non-R&D occupations. Employers report that the most important skills said to be missing from job applicants are specialist skills or knowledge. This highlights the complex and specialist nature of R&D jobs. Gaps were also found in applicants’ ability to solve complex problems and specific IT skills. Soft skills that require improvement among R&D staff include time management, task prioritisation, and teamwork. These trends in skill gaps were also evident in the literature. This highlighted the need for upskilling and reskilling for positive economic returns and a productivity uplift. The results emphasise the clear difficulties R&D employers face in filling their vacancies with adequately skilled workers, suggesting interventions may be necessary to address this issue.

As with other features of the R&D workforce, difficulty filling R&D vacancies does not appear to be uniform across regions of the UK. The South East accounts for the largest share of the UK’s total R&D ‘hard-to-fill’ vacancies. However, the East Midlands has the highest proportion of hard-to-fill vacancies within R&D-intensive sectors as a share of all vacancies in that region. This suggests a high demand for R&D skills in this area. In the UK, of the hard-to-fill vacancies within R&D sectors, 92% result from skill shortages. These findings can be useful to address recruitment difficulties and can inform policies and strategies aimed at developing a higher skilled workforce. In response to the uneven distribution of skills shortage vacancies, region-specific initiatives may be most effective when directed towards professional development and upskilling. This would address skill gaps, therefore enhancing workforce readiness and supporting regions’ economic vitality more effectively. In the R&D sectors, efforts could prioritise both the development of high-level technical skills and continued investment in soft skills training. Moreover, region-specific initiatives could be directed towards strengthening the links between higher and further education providers and businesses.

It is critical to address these hard-to-fill vacancies. The analysis shows they can lead to a range of negative consequences for R&D employers, including increased workload for staff, delays in developing new products or services, difficulties meeting customer service objectives, and increased operating costs. More broadly, these consequences negatively influence productivity of R&D establishments, which, in turn, slows down economic growth.

7.4 Next steps

This report produced valuable insights into R&D workforce trends, characteristics, skills, employer and employee perspectives, and potential future projections. However, further research is, and will continue to be, necessary. This further research will shed more light on the issues identified through this work that may threaten the UK’s ability to achieve its ambitions around science, technology, and innovation. It will also help to identify opportunities to maximise the potential of the UK’s growing R&D workforce.

8. Annex A1 - Methodology

8.1 Literature review – understanding skills matches and mismatches

We performed a selective review of the literature. We drew on existing knowledge of the subject and supplemented it with targeted searches on important issues of labour and skills demand, supply issues, and the match between supply and demand. We concentrated on the academic and grey literature from the past 10 years. This included landmark government publications. We focused on more recent literature. This included the influence of COVID-19 on the issues addressed in the review. The primary focus of the review was the UK. Though, we also drew on international literature (mainly from Europe and the USA). The review comprises 3 main sections:

  1. Matches and mismatches – outlining the definition and measurement of skills; important concepts of skills mismatches, surpluses, shortages, deficiencies and skills utilisation; and aspects of job mobility – including occupational mobility, training and other mechanisms for advancement, careers guidance and geographical mobility.
  2. Employers’ perspectives – focusing on aspects of labour demand, including: the skills employers look for and how they are changing, new skills requirements, how employers respond to skills gaps, reskilling and upskilling, and the role of migration in meeting skills needs.
  3. Workers’ perspectives – focusing on aspects of labour supply, including: pipelines of graduates (particularly STEM graduates), their choices and subjects studied, equality, diversity and inclusion issues; and work-related perceptions, motivations and attitudes of employees.

8.2 Defining the R&D workforce

Defining the R&D workforce is not a straightforward process. There is no standard classification for R&D workers in the literature. This is due to the complexities of R&D activities. Many workers are likely to conduct R&D in miscellaneous contexts across different industries and professions, in at least part of their role (Belt et al., 2021[footnote 104]). As a result, depending on data availability, various studies adopt differing approaches to measure R&D employment. For example, they rely on occupational/industrial classifications or relevant functional R&D concepts outlined in the OECD Frascati Manual (OECD, 2015[footnote 105]).

According to the Frascati Manual (OECD, 2015), “research and experimental development (R&D) comprise creative and systematic work undertaken in order to increase the stock of knowledge – including knowledge of humankind, culture and society – and to devise new applications of available knowledge. […] R&D personnel in a statistical unit include all persons engaged directly in R&D, whether employed by the statistical unit or external contributors fully integrated into the statistical unit’s R&D activities, as well as those providing direct services for the R&D activities (such as R&D managers, administrators, technicians and clerical staff).” Based on the Frascati definition, R&D personnel comprise “highly trained scientists and engineers (researchers), technicians with high levels of technical experience and training, and supporting staff who contribute directly to carrying out R&D projects and activities in R&D-performing statistical units”.

An activity can be categorised as R&D if it:

a) targets new findings (for example, it is novel)
b) relies on new concepts (creative)
c) entails uncertainty about the final result (uncertain)
d) is performed using planned methods and specific sources of funding (systematic)
e) generates a result and new knowledge that is transferable/repeatable

The Office for National Statistics (ONS) developed Standard Occupational Classification (SOC) codes. These codes provide a standardised method of defining the workforce. SOC codes classify occupations into groups at a series of digit-levels. 4-digit SOC codes denote occupations at the unit group level. These SOC code classifications are based on the skills and tasks undertaken by individuals in their respective occupations. As a result, they are well suited for defining the R&D workforce based off participation in R&D tasks and activities. Furthermore, organisations undertaking labour market analysis use these codes widely, and they are present in a range of datasets.

This study uses the Office for National Statistics (ONS) Standard Occupational Classification (SOC) to define the R&D workforce. In particular, the term ‘R&D workforce’ adopted throughout this report refers to people whose main job is denoted by specific 4-digit SOC codes. These have been diligently categorised as ‘R&D’ based on the following criteria.

First, we reviewed the task description and relevant job titles of the SOC codes. We selected occupations that are more likely to involve at least one of the following activities[footnote 106]. These activities are an extension of the subcomponents of R&D, as recommended in the Frascati Manual:

  • Basic research: work to gain new knowledge without a specific application
  • Applied research: work to obtain new knowledge relating to a specific goal or objective
  • Experimental development: systematic work that uses existing knowledge to substantially improve or create new products, services or processes
  • Market research of potential users, marketing or competitors
  • R&D or innovation management: establishing strategies, processes, structures, and responsibilities to increase research or innovation and its usage

Second, we focus only on the SOC codes related to highly skilled employment. This includes occupations within the top 3 broad categories of the SOC. The headings of these categories are:

  • ‘Managers, directors and senior officials’
  • ‘Professional occupations’
  • ‘Associate professionals’

Third, we include technicians associated with the professional occupations of our R&D-related list of SOC codes. More specifically, if a professional occupation is embodied in the R&D workforce, we also incorporate the corresponding technical occupation into our list (for example, ‘civil engineers’ and ‘building & civil engineering technicians’). In doing so, we recognise the vital role of technical staff and skills in the research and innovation sector (see relevant evidence in The TALENT commission, 2022[footnote 107]).

Fourth, we generally treat all occupations within the same broader 3-digit SOC group identically. For instance, from the SOC 2010 group ‘212 – Engineering professionals’, we include all engineers in our definition of the R&D workforce.

Fifth, we exclude ‘media professionals’ and ‘artistic, literary and media occupations’ from our R&D workforce definition. According to our approach, people employed in these occupations seem less likely to engage in research or experimental work. Similarly, we opt to omit finance-related professions from our analysis, as they do not fit in well with the R&D activities described above.

Given the complexity concerning the definition of the R&D workforce (Belt et al., 2021[footnote 108]), we created 2 groupings of R&D workers. The intention is to offer an extended and relatively comprehensive analysis. The primary focus throughout this report is our ‘core occupations’ definition of R&D employment. This encompasses 28 occupations based on the 4-digit level of SOC 2010 (referring to workers’ main job), which relate to the following broader groups of professions:

  • Science, research, engineering, and technology professionals
  • Higher education teaching professionals
  • Selected business research professionals (for example, actuaries, economists, statisticians, and other business and related research professionals)
  • Technicians who are associated with the above professions

R&D occupations given by SOC codes in the core definition of the R&D workforce

SOC 2010 Occupation title SOC 2010 Definition
2111 Chemical scientists Core R&D
2112 Biological scientists and biochemists Core R&D
2113 Physical scientists Core R&D
2114 Social and humanities scientists Core R&D
2119 Natural and social science professionals n.e.c. Core R&D
2121 Civil engineers Core R&D
2122 Mechanical engineers Core R&D
2123 Electrical engineers Core R&D
2124 Electronics engineers Core R&D
2126 Design and development engineers Core R&D
2127 Production and process engineers Core R&D
2129 Engineering professionals n.e.c. Core R&D
2135 IT business analysts, architects and systems designers Core R&D
2136 Programmers and software development professionals Core R&D
2139 Information technology and telecommunications professionals n.e.c. Core R&D
2150 Research and development managers Core R&D
2311 Higher education teaching professionals Core R&D
2425 Actuaries, economists and statisticians Core R&D
2426 Business and related research professionals Core R&D
2429 Business, research and administrative professionals n.e.c. Core R&D
2461 Quality control and planning engineers Core R&D
3111 Laboratory technicians Core R&D
3112 Electrical and electronics technicians Core R&D
3113 Engineering technicians Core R&D
3114 Building and civil engineering technicians Core R&D
3115 Quality assurance technicians Core R&D
3116 Planning, process and production technicians Core R&D
3119 Science, engineering and production technicians n.e.c. Core R&D

This list of 28 occupations makes up the core definition of R&D for this project. In addition to the above occupations, the following occupations involve R&D in their task descriptions but were included in the broader definition of the R&D workforce.

R&D occupations given by SOC codes in the broad definition of the R&D workforce

SOC 2010 Occupation title SOC 2010 Definition
1115 Chief executives and senior officials Broader R&D
1121 Production managers and directors in manufacturing Broader R&D
1122 Production managers and directors in construction Broader R&D
1123 Production managers and directors in mining and energy Broader R&D
1136 Information technology and telecommunications directors Broader R&D
2133 IT specialist managers Broader R&D
2134 IT project and programme managers Broader R&D
2137 Web design and development professionals Broader R&D
2141 Conservation professionals Broader R&D
2142 Environment professionals Broader R&D
2211 Medical practitioners Broader R&D
2212 Psychologists Broader R&D
2213 Pharmacists Broader R&D
2214 Ophthalmic opticians Broader R&D
2215 Dental practitioners Broader R&D
2216 Veterinarians Broader R&D
2217 Medical radiographers Broader R&D
2218 Podiatrists Broader R&D
2219 Health professionals n.e.c. Broader R&D
2423 Management consultants and business analysts Broader R&D
3131 IT operations technicians Broader R&D
3132 IT user support technicians Broader R&D

The analysis of Lightcast job vacancy data specifically focuses largely on the core definition of the R&D workforce. Job postings in the core definition are more likely to be specifically R&D activity-performing roles. However, at times, the analysis brings in occupations in the broader definition for more comprehensive insight. As discussed, these SOC codes are classified based on tasks and skills. As much of this work focuses on skill demand, it can be beneficial to group closely related R&D occupations. This can be accomplished by grouping the 4-digit codes into their 3-digit counterparts, with the exception of ‘2461 – Quality control and planning engineers’. These are grouped with other engineering occupations in the definition. ‘2311 – Higher education teaching professionals’ and ‘2150 – Research and development managers’ do not fit into any grouping, so they form their own groups. To avoid grouping potentially less R&D-relevant data with data for occupations in the core definition, broader R&D codes were not grouped. See the table below for details on these groupings.

Core definition SOC codes grouped into 7 corresponding ‘Occupation groups’

Occupation group SOC codes in the group
IT/programmers 2135, 2136, 2139
Engineers 2121, 2122, 2123, 2124, 2126, 2127, 2129, 2461
Technicians 3111, 3112, 3113, 3114, 3115, 3116, 3119
Scientists 2111, 2112, 2113, 2114 ,2119
Business research/analysts 2425, 2426, 2429
HE teaching professionals 2311
R&D managers 2150

8.3 Estimating the size of the R&D workforce

The Office for National Statistics (ONS) estimated and published[footnote 109] the total number of people employed in R&D occupations (as defined in Section 8.2). They used weighted data from each year of the Annual Population Survey from 2011 to 2023, based on selected SOC 2010 codes[footnote 110]. Figures are reported in Section 3.2. This differs from the approach taken to understand demographic breakdowns of the R&D workforce described in Section 8.4, where data is pooled for multiple years and unweighted. It also differs from the approach taken for projecting the future workforce size described in Section 8.8, where SOC 2020 is used to identify roles essential for R&D.

8.4 Annual Population Survey analysis: Approach and techniques

This component of the study uses the ONS Annual Population Survey (APS)[footnote 111] data from July 2012 to June 2021 to:

  • portray the profile of the R&D workforce by socio-demographic and other characteristics
  • explore differences in the workers’ likelihood of engaging in R&D activities across UK regions and over time[footnote 112]

The APS datasets gather individual-level information from the corresponding Quarterly Labour Force Surveys (QLFS) in the UK.

The sample used in this analysis includes individuals aged 16 to 65 who are employed in the UK. To increase the sample size, the APS data for 2012 to 2021 is pooled. This results in approximately 739,000 workers. 41,800 or 84,000 of these workers in the R&D workforce are based on ‘core’ and ‘broader’ definitions, respectively. This allows us to compare the R&D sample to either 655,000 or 612,000 non-R&D workers (depending on the R&D definition used).

The study descriptively explores a wide range of variables known to influence occupational choices based on relevant theoretical frameworks and empirical evidence (Blau et al., 1956[footnote 113]; Dolton et al., 1989[footnote 114]; Drost, 2002[footnote 115]; Xu, 2012[footnote 116]; Holmes et al., 2017[footnote 117]; Speer, 2017[footnote 118]; Lent et al., 2018[footnote 119]). This allows the study to explore regional variations in the likelihood of being an R&D worker.The variables used in the study can be grouped into 3 categories:

  • demographic characteristics
  • education characteristics
  • occupation/sector characteristics

The demographic characteristics include variables such as:

  • gender
  • ethnicity
  • nationality
  • age
  • region of residence and workplace
  • marital status
  • presence of children in the household
  • long-standing health issues

The education characteristics encompass variables related to:

  • the level of highest qualification
  • degree-level qualifications
  • subject of study
  • ‘good degree’ (capturing workers with a first or upper-second class degree)
  • type of university attended (Russell Group universities and other institutions)

The occupation/sector characteristics involve variables related to:

  • industry section
  • employment status (employee or self-employed)
  • workplace size
  • number of hours worked per week (inclusive of overtime)
  • employment type (full time or part time)
  • job contract (permanent or fixed-time agreement)
  • public or private sector employment
  • years with the current employer

The study presents information on ‘gross hourly wages’ for both R&D and non-R&D workers. There are adjustments for inflation and trimming of extreme values to reduce the influence of outlying observations and measurement errors.

We use logistic regression models to estimate the likelihood of individuals being employed in R&D jobs. This is in addition to the thorough descriptive analysis and after accounting for the influence of various factors outlined above. (See Section 10.4 for further details on the model specification). This econometric analysis uses 3 different models to examine how specific variables affect the probability of being an R&D worker:

  • the first model includes demographic and geographic variables
  • the second model adds education, occupation, and sector characteristics
  • the third model explores interaction effects between workplace region and other variables to account for potential differences across regions

8.5 R&D skills demand assessment via online job posting data

This section of the study aims to improve our understanding of the demand for different occupations and skills in the UK’s R&D workforce, including how this demand has varied over time and across different regions of the UK. This allows for a deeper look into the in-demand skills for R&D employers.

This analysis uses Lightcast job vacancy data for the years 2013 to 2022. This was compiled by scraping millions of online job postings to create a large database of vacancies over a period of a decade. The data includes a wealth of information relating to job titles, occupations, salaries, spatial variables, and employer-related fields. It provides an up-to-date picture of workforce and skill demand in the UK. The analysis focuses on R&D roles identified through 4-digit SOC codes. For the purpose of the investigation, demand was defined as the frequency with which a skill or occupation appears in the dataset as a proportion of the total frequency. The Lightcast taxonomy for skills groups skills at different levels of aggregation. This allows us to identify the demand for R&D-specific skills, as well as occupation-specific skills, within our R&D definition:

Term Description
Baseline skill General skills that are not specific to a role, such as ‘teamwork’ or ‘communication’. May be referred to interchangeably as ‘general skills’ throughout this report.
Specialised Skill Skills that are more specific to a role or occupation, and therefore require more specialist training or education, such as ‘Python’ or ‘clinical research’. These are mutually exclusive with baseline skills, so a skill cannot be categorised as both.
Software Skill Skills related to use of any computer software, such as ‘Microsoft Excel’. Unlike baseline and specialised skills, which are mutually exclusive, software skills are also classified as either baseline or specialised skills
Skill Cluster Family A broad category of skills such as ‘Information Technology’ and ‘Finance’.
Skill Cluster A more granular grouping than skill cluster families, such as ‘Scripting Languages’.
Skill Individual skills at the most granular level, pulled out from the job text, such as ‘Python’ or ‘Communication’.

8.6 Approach to analysing Employer Skills Survey data

Assessing skills mismatches presents significant challenges. Hence, various metrics have been put forward in the context of labour market research to gauge such mismatches (see Cardenas Rubio, 2020[footnote 120], for a comprehensive discussion). The UK Migration Advisory Committee (MAC) has divided labour shortage indicators into 4 categories:

  • employer-centred
  • price-centred
  • volume-centred
  • labour market imbalance indicators ([footnote 121])

This study draws on metrics based on employer input. These metrics originate from surveys that probe employers about their skills needs and their success in sourcing employees. They provide an insightful perspective from the employer’s standpoint. However, they are limited by the sole reliance on voluntary employer disclosures. It is also important to note that the ability of employers to identify skills shortages and gaps is likely to vary. Those at the frontier of innovation may be more likely to identify shortages and gaps.

This section of the study uses data from the Employer Skills Survey (ESS)[footnote 122] from 2019. (This was the latest available data at the time of analysis.) It looks at a wide range of skills-related issues that businesses face concerning both their existing personnel and incoming staff. It also looks at methods employed to deal with these issues. The survey includes responses from over 81,000 establishments across various sectors and considers the influence of skills shortages on operational efficiency. The analysis distinguishes between R&D-intensive sectors and others. This distinction is based on the ratio of business R&D expenditure to gross value added (GVA). ‘R&D sectors’ are defined as industries with high or medium-high R&D intensity according to the OECD’s classification. There is a slight adjustment to include lower-ranked industries with a comparably high R&D intensity specifically in the UK (Galindo-Rueda and Verger, 2016[footnote 123]; Carvalho, 2022[footnote 124]). The final selection of industries that constitute ‘R&D sectors’ consists of 86 codes from the UK Standard Industrial Classification (SIC 2007) based on the granularity of the 4-digit level.

The study uses weighted results to represent the entire population of establishments in UK regions and nations. This approach is consistent with the Department for Education’s Employer Skills Survey 2019 report (Winterbotham et al., 2020[footnote 125]).

The fieldwork for the ESS 2019 was completed prior to the COVID-19 pandemic in early 2020. This unexpected shock resulted in significant changes in the economic landscape post-survey. The present analysis does not capture these changes. Also, employers based in Scotland were not included in the ESS 2019. The findings cover England, Wales and Northern Ireland only.

Definitions of important concepts and metrics used are provided below. These definitions ensure a comprehensive understanding of the findings and analyses in this section of the study.

Concept Description
Hard-to-fill vacancies (HFV) Vacancies that employers struggle to fill because of difficulties including, but not limited to, the low number of applicants with the required skills/experience/qualifications, poor recruitment channels, poor terms and conditions offered (for example, salary), not enough people doing the required type of job, undesirable hours, difficulty with work permits/immigration issues, poor career progression, and so on.
Skills-shortage vacancies (SSV) A specific type of hard-to-fill vacancies that occur due to applicants lacking the skills, experience or qualifications employers require.
Skills gaps The total number of staff that lack full proficiency.
Incidence of vacancies The proportion of establishments reporting at least one vacancy.
Incidence of hard-to-fill vacancies The proportion of establishments reporting at least one hard-to-fill vacancy.
Incidence of skills-shortage vacancies The proportion of establishments reporting at least one skills-shortage vacancy.
Incidence of skills gaps The proportion of establishments that reported any of their staff lacked full proficiency.
Density of vacancies The number of vacancies as a proportion of all employment.
Density of hard-to-fill vacancies The number of HFV as a proportion of all vacancies.
Density of skills-shortage vacancies The number of SSV as a proportion of all vacancies.
Density of skills gaps The number of skills gaps as a proportion of all employment.

The qualitative analysis explores the underlying behaviours and perceptions of firms and individuals. It offers insights that quantitative methods cannot capture. This qualitative dimension enriches our understanding. It reveals the ‘why’ behind actions and attitudes in the R&D sectors. It is noted that the qualitative findings are not confined to a single section. Instead, they are interwoven throughout the present report. This provides contextual depth and personal perspectives to complement the quantitative data.

This work is based on 22 semi-structured qualitative interviews. The interviews were conducted between August and October 2022 with R&D employees, employers and representatives from organisations supporting, or with an interest in, R&D skills development in the West Midlands. Semi-structured interviews were chosen. This method provides sufficient flexibility regarding possible perspectives and outcomes. It also allows the interviewers to ask in-depth follow-up questions. Interview transcripts were analysed thematically. This method was selected to:

  • allow a systematic review of a large volume of qualitative data
  • identify important themes and explore important issues in more detail

The box below summarises the important research questions that the qualitative analysis answers:

1. What are the limiting factors in expanding the skills needed for R&D, research commercialisation, and technology adoption?

a. What challenges do employers face in recruiting and retaining R&D workers?
b. How can these issues be addressed to help employers find the people with R&D skills that they need?

2. What are the barriers to participation/progression in the R&D workforce for under-represented groups (for example, women, specific ethnic minorities, and disabled people in STEM occupations)? How can these barriers be overcome?

3. What are the primary developments/issues in the education & training systems concerning the supply of R&D skills?

a. How can universities and STEM assets (for example, innovation centres, incubators, science parks) increase the provision of R&D skills and encourage knowledge transfer?
b. How connected are the approaches of higher education, further education, private providers, and innovation systems?
c. Can (degree) apprenticeships and new developments (for example, T levels) contribute to expanding the provision of R&D skills? (if so, in what ways?)

The West Midlands was selected as a case study for this research. It was selected to explore experiences in a region outside of the Greater South East (GSE) regions. The GSE regions hold the global R&D hubs of London, Oxford and Cambridge. R&D activity is concentrated in these regions. The West Midlands has a smaller share of R&D workers in the total workforce (5.5% versus a national average of 6.0%), as explained in the APS analysis of the present study. Prior to the COVID-19 pandemic, the West Midlands was the fastest growing region outside London. This was “driven in particular by a thriving automotive manufacturing base, construction and business and professional services”. The West Midlands also produced 7% of UK R&D expenditure in 2022[footnote 126]. However, the West Midlands suffered from low productivity with business productivity 16% below the national average, reflecting a bigger ‘long-tail’ of less productive firms, and fewer ‘frontier’ firms (WMCA, 2021[footnote 127].

Interview participant details

1. Manufacturing CEO​ 12. STEM institute Director​
2. Representative of a manufacturing organisation​ 13. STEM Asset Representative​
3. Large engineering company Director​ 14. STEM education representative​
4. Digital Engineering Director​ 15. University partnership representative​
5. Energy company representative​ 16. Sustainability champion​
6. Energy infrastructure organisation representative​ 17. Representative of a local non-statutory body​
7. Life Sciences Company Director​ 18. Local Economic Development Organisation representative​
8. Health technologies representative 1​ 19. Civil engineer (employee)​
9. Health technologies representative 2​ 20. Technology Transfer Engineer (employee)​
10. Representative of a regional high-tech organisation​ 21. HEI Senior Technician 1 (employee)​
11. Regional Innovation representative​ 22. HEI Senior Technician 2 (employee)​

8.8 Workforce projections – Skills Imperative 2035

This section of the study uses projection data from the Nuffield Foundation[footnote 128] funded Skills Imperative 2035 research programme. The programme’s aim is to explore forecasted trends in R&D jobs.

The Institute for Employment Research and Cambridge Econometrics constructed a series of labour market outlooks for 2035 based on different scenarios (Wilson et al., 2022c[footnote 129].

The ‘Baseline’ scenario used the existing trajectory of technological advancement in 2021, political landscapes, demographic changes, and environmental transformations as its cornerstone (Wilson et al., 2022b[footnote 130]. This scenario served as the foundation for the projected employment levels between 2021 and 2035. However, it did not predict potential policy changes that could influence new trends.

The other (‘alternative’) scenarios explored possible deviations from the ‘Baseline’ scenario. They factored in accelerated technological adoption (Wilson et al., 2022a[footnote 131]. In particular, the interim ‘Automation’ scenario looked at the fast track of Artificial Intelligence and other advanced technologies. It focused on potential job losses while disregarding any new job creation. In contrast, the 2 principal alternative scenarios recognised that technology will also generate new jobs.

Specifically, the ‘Technological opportunities’ scenario suggested that investing in tech could boost productivity. This would have the benefit of pushing towards a carbon-neutral economy and improved social services.

The ‘Human-centric’ scenario gave more weight to social services, healthcare, education, and skills less likely to be automated (such as soft skills), alongside some technological and environmental focus.

The analysis included in this section of the report concentrates on the ‘Baseline’ scenario and the ‘Technological opportunities’ and ‘Human-centric’ scenarios. The annual projection rates[footnote 132] for each of these 3 scenarios were applied to July 2023 ONS occupation employment figures for R&D SOC 2020 codes to provide refreshed projections up to 2035. In absence of alternative information, this method makes the assumption that growth rates for employment would match that of jobs projected in the Skills Imperative analysis. This aligns with the approach taken by Skills England Analysis[footnote 133].

These scenarios offer 2 potential future labour market realities by 2035. However, they are merely projections and should not be regarded as definite predictions of what will happen. Moreover, these projections were created prior to subsequent geopolitical events, such as the Russian invasion of Ukraine and the consequent macroeconomic effects (including the energy crisis). However, these developments were not expected to significantly influence structural trends across occupations and industries in the long run.

In this analysis, we adopt the same methodology as outlined in Section 8.2 for defining the R&D workforce, employing certain criteria. We use the ‘core’ definition of the R&D workforce. This is derived from the Office for National Statistics (ONS) Standard Occupational Classification (SOC). This definition yields 34 R&D-related occupations at the 4-digit level of SOC 2020, corresponding to workers’ main jobs. These encompass:

(a) professionals in the science, research, engineering, and technology fields

(b) higher education teaching professionals

(c) selected business research professionals (including actuaries, economists, statisticians, and other similar roles)

(d) technicians linked to the professions above

The method described in Section 8.2 used SOC 2010 for defining the R&D workforce. For our projections analysis, we applied the same selection criteria to identify R&D-related occupations using the available projection data, which are based on SOC 2020.

The Skills Imperative projections assume that the share of employment across 4-digit SOC codes within the 2-digit class will remain constant over the projected period. Each 4-digit SOC within each class is assumed to have the same growth rate. This approach is taken due to a lack of available data at the 4-digit SOC level. However, it is highly possible that growth rates of occupations at the 4-digit level could vary within each group. Therefore, projections at the 4-digit level should not be interpreted as precise or certain predictions.

 9. Annex A2 – Supplementary figures

Figure A1. Proportion of the workforce working in R&D occupations (‘core’ definition) by region of residence (2012 to 2021

Source: APS datasets (July 2012 to June 2021) unweighted, authors’ calculations

Figure A2. Probability of being an R&D worker (‘core’ definition) by gender (average adjusted predictions)

Note: The graph shows the predicted likelihood of being an R&D worker conditional on the variables included in the extended logistic regression model (Model 2 of Table A10 in the accompanying Annex data tables). The shaded areas depict the 95% confidence intervals of the average adjusted likelihood (unlike other similar graphs in the report, these intervals are too narrow to be clearly visible due to the large sample size for women and men). Source: APS pooled datasets (July 2012 to June 2021) unweighted, authors’ calculations

Figure A3. Probability of being an R&D worker (‘core’ definition) by ethnicity (average adjusted predictions)

Note: The graph shows the predicted likelihood of being an R&D worker conditional on the variables included in the extended logistic regression model (Model 2 of Table A10 in the accompanying Annex data tables). The shaded areas depict the 95% confidence intervals of the average adjusted likelihood. Source: APS pooled datasets (July 2012 to June 2021) unweighted, authors’ calculations

Figure A4. Probability of being an R&D worker (‘core’ definition) by nationality (citizenship) (average adjusted predictions) |

Note: The graph shows the predicted likelihood of being an R&D worker conditional on the variables included in the extended logistic regression model (Model 2 of Table A10 in the accompanying Annex data tables). The shaded areas depict the 95% confidence intervals of the average adjusted likelihood. Source: APS pooled datasets (July 2012 to June 2021) unweighted, authors’ calculations.

Figure A5. Probability of being an R&D worker (‘core’ definition) by region of workplace (average adjusted predictions)

Note: The graph shows the predicted likelihood of being an R&D worker conditional on the variables included in the extended logistic regression model (Model 2 of Table A10 in the accompanying Annex data tables). The shaded areas depict the 95% confidence intervals of the average adjusted likelihood. Source: APS pooled datasets (July 2012 to June 2021) unweighted, authors’ calculations

Figure A6. Distribution of R&D establishments by 2-digit SIC 2007 industry

Note: The figures in the graph are weighted based on the number of establishments. The industries ‘Manufacture of fabricated metal products, except machinery & equipment’ (SIC 2007 code: 25), ‘Publishing activities’ (SIC 2007 code: 58) and ‘Manufacture of basic pharmaceutical products & pharmaceutical preparations’ (SIC 2007 code: 21) are not reported because of the small size of the underlying unweighted cell counts. These 3 unreported industries cover 0.7% of the total number of R&D establishments. Figures refer to England, Wales, and N. Ireland (excluding Scotland). Source: Employer Skills Survey 2019, authors’ own calculations

Figure A7. Most prevalent IT skills difficult to obtain from applicants by sector group

Note: The graph shows the weighted proportion of establishments reporting a particular skill shortage. Base: Establishments with skills-shortage vacancies caused by a lack of IT skills. Figures refer to England, Wales, and N. Ireland (excluding Scotland). Source: Employer Skills Survey 2019, authors’ own calculations

Figure A8. Impact of hard-to-fill vacancies on employers by sector group

Note: The graph shows the weighted proportion of establishments reporting a particular impact. Base: Establishments with hard-to-fill vacancies. Figures refer to England, Wales, and N. Ireland (excluding Scotland). Source: Employer Skills Survey 2019, authors’ own calculations

Figure A9. Actions taken by employers to overcome hard-to-fill vacancies by sector group

Note: The graph shows the weighted proportion of establishments reporting a particular action. Base: Establishments with hard-to-fill vacancies. Figures refer to England, Wales, and N. Ireland (excluding Scotland). Source: Employer Skills Survey 2019, authors’ own calculations

Figure A10. Most common technical/practical skills that need improving among staff by sector group

Note: The graph shows the weighted proportion of establishments reporting a particular skill gap. Base: Establishments with skills gaps. Figures refer to England, Wales, and N. Ireland (excluding Scotland). Source: Employer Skills Survey 2019, authors’ own calculations

Figure A11. Most common IT skills that need improving among staff by sector group

Note: The graph shows the weighted proportion of establishments reporting a particular skill gap. Base: Establishments with skills gaps related to a lack of IT skills. Figures refer to England, Wales, and N. Ireland (excluding Scotland). Source: Employer Skills Survey 2019, authors’ own calculations

Figure A12. Soft/people skills that need improving among staff by sector group

Note: The graph shows the weighted proportion of establishments reporting a particular skill gap. Base: Establishments with skills gaps. Figures refer to England, Wales, and N. Ireland (excluding Scotland). Source: Employer Skills Survey 2019, authors’ own calculations

Figure A13. Actions taken by employers to overcome skills gaps among staff by sector group

Note: The graph shows the weighted proportion of establishments reporting a particular action. Base: Establishments with skills gaps. Figures refer to England, Wales, and N. Ireland (excluding Scotland). Source: Employer Skills Survey 2019, authors’ own calculations

10. Annex B –Technical annex

10.1 Skills mismatches: detailed methodology

Background on skills mismatches

A ‘skills mismatch’ occurs when skills supply does not meet skills demand. This occurs when either supply is less than demand (the top left cell in Figure B1) or when supply is greater than demand (the bottom right box in Figure B1). Skills mismatches can constitute economic growth. There are several concepts related to skills mismatch.

Figure B1: Inter-relationship between low and high levels of skills supply and demand

Source: Adapted from Green A. (2016 [footnote 134])

A ‘skills shortage’ occurs when supply is less than demand. Typically, this is measured by skills shortage vacancies on the external labour market. However, data on skills shortage vacancies from employer surveys is limited. They only provide what employers choose to report. When some employers speak of skills shortages, they may be referring to the social skills that they require to fit into their workplace (Green and Ashton, 1992[footnote 135]). This has been referred to as the ‘good bloke syndrome’ (Oliver and Turton, 1982[footnote 136]). This refers to the ability to work in a manner consistent with managerial interests, including prevailing wage rates and ways of working. Greater than average wage growth is one measure used in identifying skills shortages (Gambin et al., 2016[footnote 137]), including by the Migration Advisory Committee (2017)[footnote 138] in identifying ‘shortage occupations’. In the case of a labour shortage, market pressure should increase wages. This helps to raise supply and reduce demand and restore labour market equilibrium. On this basis, rising wages within an occupation can be considered to provide an indication of a shortage. 

Another concept related to supply being less than demand is a ‘skills gap’. This occurs on the internal labour market when managers assess that employees do not meet the competence levels required for a particular job role.  

A related concept is ‘undereducation’. This is when the education level of workers and other individuals in the available talent pool (such as job seekers) is less than what is required for certain jobs or industries. Some skills shortages would be expected in a healthy and dynamic economy as employers raise their product market ambitions.

A ‘skills surplus’ occurs when supply is greater than demand. Typically, this is measured by the unemployment rate (for example, when there is a surplus of individuals with appropriate skills on the external labour market). ‘Skills underutilisation’ is when workers are employed below their skills level. This is a difficult concept to measure, given that it usually relies on subjective indicators in surveys of employees. Answers can say more about self-efficacy and self-confidence than about use of skills (Green and Ashton, 1992). A related, but indirect, concept is ‘overeducation’. This is when workers’ education level is greater than is required.  

The issue of ‘Skills underutilisation’ can be associated with lower pay, job satisfaction and productivity (Allen and van der Velden, 2001[footnote 139]; McGuiness and Sloane, 2011[footnote 140]; Kampelmann and Rycx, 2012[footnote 141]). Recent research shows that by comparison with matched graduates in task-warranted (‘typical’) graduate jobs, graduates in task-unwarranted graduate jobs and in non-graduate jobs perceive lower skills utilisation and experience negative wage gaps (Green and Henseke, 2021[footnote 142]). More generally, research emphasises that ‘management quality’ and the broader institutional culture are central to shaping the opportunities made available to, and the incentives for, employees to use their skills in the workplace. Practices for more effective skills utilisation include job re-design, job rotation and multi-skilling. Skills that are not practised can deteriorate. However, some workers choose job roles below their skills level, often for non-work reasons (Lyons et al., 2020[footnote 143]). 

Skills are in equilibrium when supply matches demand. A ‘high skills equilibrium’ occurs when the economy demands high level skills in high wage high productivity ‘good jobs’ and the supply of labour meets demand. The policy challenge is to maintain this position dynamically as demand changes. A ‘low skills equilibrium’ occurs when demand for skills is low and so there is a lack of incentives for employers and individuals to invest in skills. 

A ‘skills deficit’ can occur when the supply and demand for skills are in equilibrium but at a sub-optimal (or feasible) level. This is the case of a low skills equilibrium or in instances of equilibrium at medium or higher skills levels. A skills deficit is difficult to measure because it implies comparison with a specific optimum. For example, a benchmark of a (similar) country or region that is characterised by a greater supply or demand for skills. It has been argued that skills deficits should have as much prominence in debates on skills issues as skills shortages (Green F., 2016). Policy action to alleviate skills deficits is needed to:

  • influence both demand for and supply of skills
  • encourage employers to raise demand for skills
  • encourage individuals to invest in their human capital

Data 

To address its research objectives, the present study uses data from the Employer Skills Survey1 conducted in 2019. The ESS proves instrumental in identifying a range of skills-related issues that businesses confront, concerning both their existing personnel and incoming staff. Furthermore, it unveils the methods these businesses employ to address these problems.  

The survey’s 2019 version incorporated responses from a large number of employers. This made it one of the most exhaustive global business surveys (Winterbotham et al., 2020). The survey focused on establishments across all sectors in England, Northern Ireland, and Wales, including businesses with a minimum of 2 paid employees. Unlike the firm-level, the establishment level of the analysis acknowledges the localised influence of labour markets and skills issues. It recognises that skills-related challenges are often experienced more intensely at each work site, rather than at the encompassing organisational level (Winterbotham et al., 2020). The ESS 2019 relies on the responses of over 81,000 establishments. It allows an assessment of the effects of skills shortages on the operational efficiency of companies and organisations at various scales – from national to local, and even to sectoral. It records the prevalence, type, and consequence of these skills-related issues.

Definitions

The present analysis distinguishes between R&D-intensive sectors and others. R&D intensity is defined as the ratio of a specific industry’s business R&D expenditure to its gross value added (GVA). We obtained information about the R&D intensity of each industry from the OECD’s taxonomy of economic activities (Galindo-Rueda and Verger, 2016[footnote 144]). This taxonomy classifies industries into 5 major groups according to their level of R&D intensity (high, medium-high, medium, medium-low, and low). In the current setting, we adopted the proposed OECD’s classification and selected industries with ‘high’ or ‘medium-high’ R&D intensity to define ‘R&D sectors’.  

To accommodate UK-specific industrial features, we slightly broadened the definition of ‘R&D sectors’. The revised definition encompasses any other UK industries where the R&D intensity surpasses that of the industry classified as having the lowest R&D intensity among those with ‘high’ or ‘medium-high’ intensity, as per the OECD taxonomy mentioned above. To achieve this, we referenced a recent report by the Enterprise Research Centre (Carvalho, 2022[footnote 145]). The only industry that meets the latter criterion pertains to the construction of ships and boats. The final selection of industries that constitute ‘R&D sectors’ comprises 86 codes from the UK Standard Industrial Classification (SIC 2007) based on the granularity of the 4-digit level (see Table below).  

There were some limitations stemming from the use of SIC codes to identify innovative firms and the increasing popularity of new techniques for classifying industrial activities. These included text mining algorithms (Marra and Baldassari, 2022[footnote 146]). The method employed here offers a valuable insight into the challenges faced by R&D-intensive firms, given the variables available in the ESS datasets. 

10.2 Industries comprising the ‘R&D Sectors’ by 4-digit SIC codes

R&D SIC codes and their descriptions

SIC 2007 class Description
20.11 Manufacture of industrial gases
20.12 Manufacture of dyes and pigments
20.13 Manufacture of other inorganic basic chemicals
20.14 Manufacture of other organic basic chemicals
20.15 Manufacture of fertilisers and nitrogen compounds
20.16 Manufacture of plastics in primary forms
20.17 Manufacture of synthetic rubber in primary forms
20.20 Manufacture of pesticides and other agrochemical products
20.30 Manufacture of paints, varnishes and similar coatings, printing ink and mastics
20.41 Manufacture of soap and detergents, cleaning and polishing preparations
20.42 Manufacture of perfumes and toilet preparations
20.51 Manufacture of explosives
20.52 Manufacture of glues
20.53 Manufacture of essential oils
20.59 Manufacture of other chemical products n.e.c.
20.60 Manufacture of man-made fibres
21.10 Manufacture of basic pharmaceutical products
21.20 Manufacture of pharmaceutical preparations
25.40 Manufacture of weapons and ammunition
26.11 Manufacture of electronic components
26.12 Manufacture of loaded electronic boards
26.20 Manufacture of computers and peripheral equipment
26.30 Manufacture of communication equipment
26.40 Manufacture of consumer electronics
26.51 Manufacture of instruments and appliances for measuring, testing and navigation
26.52 Manufacture of watches and clocks
26.60 Manufacture of irradiation, electromedical and electrotherapeutic equipment
26.70 Manufacture of optical instruments and photographic equipment
26.80 Manufacture of magnetic and optical media
27.11 Manufacture of electric motors, generators and transformers
27.12 Manufacture of electricity distribution and control apparatus
27.20 Manufacture of batteries and accumulators
27.31 Manufacture of fibre optic cables
27.32 Manufacture of other electronic and electric wires and cables
27.33 Manufacture of wiring devices
27.40 Manufacture of electric lighting equipment
27.51 Manufacture of electric domestic appliances
27.52 Manufacture of non-electric domestic appliances
27.90 Manufacture of other electrical equipment
28.11 Manufacture of engines and turbines, except aircraft, vehicle and cycle engines
28.12 Manufacture of fluid power equipment
28.13 Manufacture of other pumps and compressors
28.14 Manufacture of other taps and valves
28.15 Manufacture of bearings, gears, gearing and driving elements
28.21 Manufacture of ovens, furnaces and furnace burners
28.22 Manufacture of lifting and handling equipment
28.23 Manufacture of office machinery and equipment (except computers and peripheral equipment)
28.24 Manufacture of power-driven hand tools
28.25 Manufacture of non-domestic cooling and ventilation equipment
28.29 Manufacture of other general-purpose machinery n.e.c.
28.30 Manufacture of agricultural and forestry machinery
28.41 Manufacture of metal forming machinery
28.49 Manufacture of other machine tools
28.91 Manufacture of machinery for metallurgy
28.92 Manufacture of machinery for mining, quarrying and construction
28.93 Manufacture of machinery for food, beverage and tobacco processing
28.94 Manufacture of machinery for textile, apparel and leather production
28.95 Manufacture of machinery for paper and paperboard production
28.96 Manufacture of plastics and rubber machinery
28.99 Manufacture of other special-purpose machinery n.e.c.
29.10 Manufacture of motor vehicles
29.20 Manufacture of bodies (coachwork) for motor vehicles; manufacture of trailers and semi-trailers
29.31 Manufacture of electrical and electronic equipment for motor vehicles
29.32 Manufacture of other parts and accessories for motor vehicles
30.11 Building of ships and floating structures
30.12 Building of pleasure and sporting boats
30.20 Manufacture of railway locomotives and rolling stock
30.30 Manufacture of air and spacecraft and related machinery
30.40 Manufacture of military fighting vehicles
30.91 Manufacture of motorcycles
30.92 Manufacture of bicycles and invalid carriages
30.99 Manufacture of other transport equipment n.e.c.
32.50 Manufacture of medical and dental instruments and supplies
58.21 Publishing of computer games
58.29 Other software publishing
62.01 Computer programming activities
62.02 Computer consultancy activities
62.03 Computer facilities management activities
62.09 Other information technology and computer service activities
63.11 Data processing, hosting and related activities
63.12 Web portals
63.91 News agency activities
63.99 Other information service activities n.e.c.
72.11 Research and experimental development on biotechnology
72.19 Other research and experimental development on natural sciences and engineering
72.20 Research and experimental development on social sciences and humanities

10.3 Lightcast analysis detailed methodology

Data

To build our understanding of demand for different occupations and skills in the UK’s R&D workforce, it is necessary to understand:

  • what employers are looking for
  • how this has varied over time
  • how this demand varies across different regions of the UK to get a clear picture of trends and likely future needs.

To achieve this, the dataset used for the analysis is Lightcast job postings data. Lightcast scrapes thousands of sources online, including job boards, job posting aggregators, recruiters and company websites. This has resulted in a large database of job postings that spans over 10 years. It includes a range of data fields including job titles, salaries, spatial variables, and employer-related fields such as industry code. The data is frequently updated, providing one of the most up-to-date pictures of workforce and skill demand available. This analysis uses data spanning from 2013 to 2022. 2022 is the latest full year of data at the time of analysis.

Occupations

One of the benefits of utilising Lightcast job postings data is that each job posting is assigned a 4-digit SOC code corresponding to the occupation of the role being advertised. This enables job postings for R&D roles to be identified and picked out for analysis. It also allows non-R&D job postings to be picked out for comparative analysis. With regards to representation of different occupations in the data, a review of Lightcast job postings data by the OECD shows that the data has overall good statistical properties, providing useful, up-to-date information[footnote 147]. 5 countries are included in Lightcast’s web scraping in the OECD’s review. The United Kingdom is the country that shows the best coverage of Lightcast vacancy data relative to official vacancy data. This coverage was found to remain consistently above 80% from 2012 to 2018. This suggests that the data likely provides a good representation of real trends in demand for workforce and skills in the UK. However, some occupations may be more or less represented in the data than others. (This is discussed where relevant in the report.)

Skills definitions

The ONS developed SOC codes can act as an indicator for skill demand, as codes are grouped based on skill requirements and tasks/activities. However, SOC codes alone are inadequate for the purpose of this analysis. An important aim is to gain a more in-depth understanding of specific skill demand for R&D roles, the trends in this demand, and how skill needs vary from region to region across the UK.

Lightcast job postings data facilitates this. Each job posting is linked to its associated skills, which are pulled out from individual job postings. The skills grouping contains over 24,000 unique skills. These skills are grouped at different levels. At the highest level, skills are classified as:

  • Baseline skills: general skills that are not specific to a role, such as ‘teamwork’ or ‘communication’. They can be referred to interchangeably as ‘general skills’ throughout this report
  • Specialised skills: skills that are more specific to a role or occupation. They require more specialist training or education, such as ‘Python’ or ‘clinical research’. These are mutually exclusive with baseline skills. Therefore, a skill cannot be categorised as both
  • Software skills: skills related to use of any computer software, such as ‘Microsoft Excel’. Software skills are also classified as either baseline or specialised skills

In addition to these 3 categories, skills are grouped into ‘Skill Cluster Families’. These are a broad category of skills such as ‘Information Technology’ and ‘Finance’. At a more granular level, skills are categorised into ‘Skill Clusters’ that are the lowest level aggregation of skills in the taxonomy. For the purpose of this investigation, analysis was conducted using all of these groupings, as well as at the individual skill level. This was contingent upon which level of aggregation was most appropriate to achieve the aims. Baseline skills are important to all roles. However, they are filtered out during the parts of the analysis aimed at understanding the skill profiles of R&D occupations specifically. This is because most baseline skills are nonspecific to an occupation.

Other data

There is a range of additional data fields to those discussed in the Lightcast dataset. These include the minimum level of qualification (NQF level) for a given job posting, salary data, and the particular degree qualifications specified by employers in each job posting. These additional fields are brought into the analysis where relevant to deepen our understanding of the R&D workforce and demand for skills and talent.

Caveats of the analysis

There is a potential limitation with the Lightcast dataset. The OECD review of Lightcast data states that it has good statistical properties for most years and countries. However, representation in the dataset does vary across occupations. Specifically, the review finds that ‘professional’, ‘manager’ and ‘technician’ occupations are generally overrepresented in the dataset. All of these are included in our core and broad definitions of the R&D workforce. Initial analysis of the R&D workforce demand using the data appears to support this. Demand for certain occupations in certain regions appears to be inconsistent with observed trends in the R&D workforce using Labour Force Survey (LFS) data. This discrepancy could exist for a number of reasons. One is that there is variability in participation in online job postings by occupation. Some roles may be more likely to advertise/hire internally, some may hire more informally, and so on. There is also uneven representation between sectors. Varying turnover rates by occupation can also contribute to this. This variation in representation appears to also vary over time. This can affect analysis of trends. For example, IT and programming occupations appear to decrease in relative demand over the available 10-year period. However, it may instead be the case that a greater range of other occupations are utilising online job postings over time, affecting the relative trend and the share of demand for each occupation.

Another potential limitation with the Lightcast dataset exists. While it provides us with a wealth of data on the number of job postings for R&D occupations, the skills they require, and a range of other variables, it does not tell us:

  • who applied
  • how many applicants there were
  • if the job posting was filled
  • if so, how long it took to fill the post

This information would add an important dimension to the pure demand aspect that can be pulled out of the dataset. For example, a large number of qualified applicants may point to a high supply of labour/skills meeting demand for a given occupation. A posting that takes a long time to fill (or never gets filled) can indicate the opposite.

Finally, there are some limitations with the quality of the data itself. For example:

  • allocations to SOC codes in the Lightcast dataset is automated. Therefore, complete accuracy is not guaranteed. This accuracy can also vary between occupations (skills in the dataset can have a similar problem)
  • not all online jobs are captured in the data
  • some job postings may be poor quality or duplicates

10.4 Annual Population Survey analysis – econometric framework

Logistic regression models

This study seeks to model the probability of being employed in R&D-related occupations, conditional on many demographic, education, and job/sector-pertinent characteristics. In the current setting, we consider the occupational choice a qualitative occurrence, represented by a binary outcome. Specifically, the response variable () takes 2 values (zero and one), denoting whether a person works in an R&D-related occupation or not (conditional on employment). Furthermore, the likelihood of this occurrence is associated with a large number of factors (independent variables) collected in a vector X.

One could use the linear probability model (LPM) to estimate the conditional probability of being an R&D worker. However, this approach comes with some disadvantages. Specifically, the predicted probabilities can result in unrealistic estimates in some instances (such as returning a value that is larger than one or negative). More importantly, linear regressions imply that the impact of an un-interacted explanatory variable on the likelihood of being an R&D worker is constant throughout its distribution. For example, the LPM assumes that the effect of age on the dependent variable remains unchanged at all levels of age. In other words, the ceteris paribus difference in the likelihood of conducting R&D activities between 2 workers aged 22 and 23 years would be the same as that for 2 workers aged 42 and 43, respectively. Given that the relationship between a binary outcome variable and the regressors is plausibly non-linear, the abovementioned property of the LPM would likely lead to a misspecified functional form of the model (Mood, 2009[footnote 148]).

Researchers commonly use logistic or probit models when the dependent variable is binary. Both models produce similar estimates in empirical projects, as these approaches have no substantive differences (Amemiya, 1981[footnote 149]). In this report, we adopt the logistic function to estimate the likelihood of a worker being employed in an R&D-related occupation (based on the ‘core’ definition of the R&D workforce), conditional on the other explanatory variables. Specifically, the form of the logistic regression model is:

Where:

  • Yi is the dependent variable (which equals one if the worker conducts R&D)
  • xi represents the vector of independent variables
  • B denotes the corresponding unknown population parameters that the model estimates

The report presents the econometric analysis results based on 3 separate model specifications. The idea is to illustrate how adding specific variables to the regression alters the effect the other independent variables have on the probability of being an R&D worker.

  1. Model 1 includes only demographic and geographic variables (gender, age, age squared, ethnicity, nationality, family and health characteristics, and region of workplace), employment status (employee or self-employed) and survey year dummies[footnote 150]
  2. Model 2 contains additional variables relating to education, occupation, and sector characteristics
  3. Model 3 explores potential heterogeneous effects by introducing interaction terms in the logistic regression specification

Specifically, we concentrate on the following interaction effects of interest: ‘Region of workplace x Gender’, ‘Region of workplace x Ethnicity’, ‘Region of workplace x Level of highest qualification held’, and ‘Region of workplace x Industry section’. For instance, the interaction effect between the workplace region and gender accounts for the fact that the effect of gender on the likelihood of being an R&D worker can differ across UK regions.

Average marginal effects and adjusted predictions

A direct and uncomplicated approach to reporting and interpreting the logistic regression model results is to estimate the average marginal effects (AMEs). AMEs explicitly summarise each explanatory variable’s impact on the probability of conducting R&D in a single metric, even if the regression specification includes interaction terms (Long and Freese, 2014[footnote 151]). AMEs rely on the predicted propensities estimated by the original logistic models. Taking the example of the independent variable ‘nationality’ (k= India), the AMEs are intuitively estimated in the following way (see Williams (2012[footnote 152]) for a detailed explanation). Assume that the first worker in our dataset holds Indian nationality (irrespective of his/her actual nationality). Then, conditioning on the observed values of the rest explanatory variables, estimate the predicted probability of conducting R&D for the first worker under the assumption of Indian nationality. Similarly, now consider that the first worker holds UK nationality and compute the difference between those 2 likelihoods (that is, the marginal effect of the first worker). Likewise, replicate this method for all workers in our data (n) and estimate the average marginal effects of nationality on the probability of working in R&D occupations (see equation 2).

Finally, Section 4 also graphically illustrates the average adjusted predictions (AAPs) for selected independent variables that are strongly associated with the probability of engaging in R&D activities (that is, gender, ethnicity, nationality, level of the highest qualification, region of workplace and industry section in the main job). AAPs refer to the likelihood of working in an R&D-related profession, conditional on the observed values of the other independent variables. Therefore, AAPs constitute a valuable measure for predicting the average probability of conducting R&D activities once accounting for all the observable factors incorporated in the logistic regression models.

Caveats

An individual’s choice between different possible occupations depends on many factors. These may relate to:

  • a person’s perception of the rewards offered by various jobs (for example, lifetime income and social status)
  • certain individual characteristics typically not observed in administrative data (for example, ability, motivation/aspirations, cultural capital, socio-economic background, professional networks)
  • other intraregional labour market circumstances and economic structures. For instance, individuals with high youth aspirations are more inclined to work in highly skilled jobs akin to R&D-related occupations (Schoon and Parsons, 2002[footnote 153])

The decision to follow specific career paths could be cyclical over time. This suggests that the supply of certain R&D workers during a particular period is shaped in part by the relative demand and vice versa. For example, if technology professions have more positive career prospects in a given period, this would influence the choices of young people to study relevant courses at university. Therefore, this would increase the supply of technology workers in the labour market. However, an oversupply of technology workers might consequently lead to a decrease in their average salaries. Therefore, this would affect the educational choices of students with ‘myopic’ earnings expectations in the subsequent periods (Drost, 2002[footnote 154]).

In the regression analysis of this report, due to data limitations, it is impossible to control for all the determinants that affect the occupational choices of individuals. The unobserved factors are likely to be correlated with the independent variables of interest (for example, gender, ethnicity, nationality, workplace region, level of highest qualification obtained) and the probability of being an R&D worker. If the unobserved characteristics mentioned above differ systematically across subgroups of workers, then the estimated effects of the relevant variables would suffer from omitted variable bias. For example, if ‘career aspirations’ are on average higher for a specific ethnic minority group and are also positively associated with the likelihood of entering R&D-related occupations, omitting this characteristic from the logistic regression models leads to an upward bias in the estimated impact of this ethnic group.

Similarly, the relationship between some explanatory variables incorporated in the logistic models and the likelihood of working in R&D occupations may be simultaneous, thereby introducing additional sources of bias in the estimated effects. For instance, the industry sector in a worker’s primary job affects the chosen occupation. At the same time, entering a specific profession also influences the probability of working in a particular industry sector. Finally, the selected sample contains only people in employment. This suggests that the individuals included in our analysis should be positively selected on some unobserved factors relative to those not in employment. (For example, they may represent a higher portion of the ability/motivation distribution.) Consequently, the observed differences among subgroups of workers regarding the probability of working in R&D jobs could either overstate or understate the actual population effects. For the above reasons, the results presented here should be interpreted cautiously. Although the findings represent meaningful correlations between the independent variables and the probability of conducting R&D activities, they do not establish causal relationships.

10.5 Qualification Framework

Qualification level definitions[footnote 155]

Level Example qualifications (not exhaustive)
Entry level entry level award
entry level certificate (ELC)
entry level diploma
Level 1 GCSE - grades 3, 2, 1 or grades D, E, F, G
level 1 diploma
level 1 award
Level 2 CSE - grade 1
GCSE - grades 9, 8, 7, 6, 5, 4 or grades A*, A, B, C
intermediate apprenticeship
Level 3 A level
advanced apprenticeship
T Level
Level 4 certificate of higher education (CertHE)
higher apprenticeship
higher national certificate (HNC)
Level 5 diploma of higher education (DipHE)
foundation degree
higher national diploma (HND)
Level 6 degree apprenticeship
degree with honours - for example, bachelor of the arts (BA) hons, bachelor of science (BSc) hons
graduate diploma
Level 7 integrated master’s degree, for example, master of engineering (MEng)
master’s degree, for example, master of arts (MA), master of science (MSc)
postgraduate certificate in education (PGCE)
Level 8 doctorate, for example, doctor of philosophy (PhD or DPhil)
level 8 award
level 8 certificate

Note: The National Qualifications Framework (NQF) referred to throughout this report was replaced by the Regulated Qualifications Framework (RQF) in 2015. The levels in the RQF remain the same as those in the NQF, but the RQF also defines qualifications by size, based on the estimated amount of study time.

11. Reference List

  1. DfE Employer Skills Survey, (2022) ↩

  2. Skills England to transform opportunities and drive growth ↩

  3. See Annex Section 8.2 for more information about defining the R&D workforce. ↩

  4. ONS, Employment in research and development occupations ↩

  5. Bell, A., Chetty, R., Jaravel, X., Petkova, N., and Van Reenen, J. (2018). ‘Who Becomes an Inventor in America? The Importance of Exposure to Innovation’. CEP Discussion Paper No. 1519. ↩

  6. Østergaard, C. R., Timmermans, B. & Kristinsson, K. (2011). Does a different view create something new? The effect of employee diversity on innovation. Research Policy, 40(3), pp. 500-509. ↩

  7. Ritter-Hayashi, D., Vermeulen, P. & Knoben, J. (2019). Is this a man’s world? The effect of gender diversity and gender equality on firm innovativeness. PLoS One, 14(9), pp. e0222443. ↩

  8. “Skills-shortage vacancies” are hard-to-fill vacancies that occur specifically due to applicants lacking the skills, experience or qualifications employers require. ↩

  9. Data for Scotland was unavailable in the Employer Skills Survey dataset used for this analysis. ↩

  10. “The R&D sector” is defined as industries with high or medium-high R&D intensity according to the OECD’s classification of these industries, with a slight adjustment to include lower ranked industries with a comparably high R&D intensity specifically in the UK (see the methodology and Technical Annex for more detail). ↩

  11. Hasan, I. and Tucci, C. L. (2010). The innovation–economic growth nexus: Global evidence. Research Policy, 39(10), pp. 1264-1276. ↩

  12. Maradana, R. P., Pradhan, R. P., Dash, S., Gaurav, K., Jayakumar, M. & Chatterjee, D. (2017). Does innovation promote economic growth? Evidence from European countries. Journal of Innovation and Entrepreneurship, 6(1), pp. 1. ↩

  13. Taylor, A., Nelson, J., O’Donnell, S., Davies, E. & Hillary, J. (2022). The Skills Imperative 2035: what does the literature tell us about essential skills most needed for work? Slough: NFER. ↩

  14. Green, A., Riley, R., Smith, A., Brittain, B. & Read, H. (2021). The Future Business District. WMREDI (University of Birmingham) - Colmore Business District. ↩

  15. Cardenas Rubio, J. and Hogarth, T. (2021). The R&D Pipeline. BEIS (Department for Business Energy & Industrial Strategy) - Warwick Institute of Employment Research. BEIS Research Paper Number: 2021/22. ↩

  16. De Lyon, J. and Dhingra, S. (2021). The impacts of Covid-19 and Brexit on the UK economy: early evidence in 2021. Centre for Economic Performance - London School of Economics and Political Science. ↩

  17. ONS (2021a). Changing trends and recent shortages in the labour market, UK: 2016 to 2021. ONS (Office for National Statistics). ↩

  18. OECD (2019) Skills for 2030: OECD Future of Education and Skills 2030 Conceptual learning framework Paris: OECD Publishing. ↩

  19. Industrial Strategy Council (2019). UK Skills Mismatch in 2030. London. ↩

  20. Dickerson, A. and Morris, D. (2019) ‘The Changing Demand for Skills in the UK’, Research Discussion Paper 020, CVER Discussion Paper Series - ISSN 2398-7553. ↩

  21. O’Clery, N. and Kinsella, S. (2022) ‘Modular structure in labour networks reveals skills basins’, Research Policy 51, 104486. ↩

  22. Champion, T. and Shuttleworth, I. (2017) Are People Changing Address Less? An Analysis of Migration within England and Wales, 1971–2011, by Distance of Move, Population, Space and Place 23(3) e2026. ↩

  23. Green, A. (2018) Understanding the drivers of internal migration, in Champion T., Cooke T. and Shuttleworth I. (Eds.) Internal migration in the developed world: Are we becoming less mobile?, Routledge, London, pp. 31-55. ↩

  24. Sumption, M. (2019) Is Employer Sponsorship a Good Way to Manage Labour Migration? Implications for Post-Brexit Migration Policies, National Institute Economic Review 248(1), pp. R28-R39. ↩

  25. Taylor, A. and Green, A. (2021) ‘How well equipped are national surveys to capture new approaches to training?’, Journal of Education and Work 34(5-6), pp. 676-690. ↩

  26. Green, A., Owen, D., Atfield, G., Baldauf, B., Bramley, G. & Kispeter, E. (2020). Employer decision-making around skill shortages, employee shortages and migration: Literature Review. Migration Advisory Committee. ↩

  27. Short courses that can be completed flexibly and provide certification upon completion. They are designed to be fast, accessible, and specialised. ↩

  28. Green, A. and Taylor, A. (2020). Workplace Perspectives on Skills. University of Birmingham: Birmingham. ↩

  29. Allas, T., Fairbairn, W. and Foote, E. (2020). The economic case for reskilling in the UK: How employers can thrive by boosting workers’ skills. McKinsey & Company. ↩

  30. Dondi, M., Klier J., Panier, F. and Schubert, J. (2021). Defining the skills citizens will need in the future of work. ↩

  31. Industrial Strategy Council (2019). UK Skills Mismatch in 2030. London. ↩

  32. Li, Q., Valero, A. and Ventura, G. (2020). Trends in job-related training and policies for building future skills into the recovery. Centre for Vocational Education Research. ↩

  33. Walker, T. (2020). Skills For Net-Zero in Lancashire Building The Low Carbon Workforce of the Future. Work Foundation and Lancashire Enterprise Partnership. ↩

  34. Kapetaniou, C. and McIvor, C. (2020). Going Green: Preparing the UK workforce for the transition to a net-zero economy. Future Fit and Nesta. ↩

  35. Allas, T., Fairbairn, W. and Foote, E. (2020). The economic case for reskilling in the UK: How employers can thrive by boosting workers’ skills. McKinsey & Company ↩

  36. ECITB (2020). Igniting the Spark? Apprenticeships in the Engineering Construction Industry. ↩

  37. CIPD. (2021). Addressing Skills and Labour Shortages Post-Brexit ↩

  38. McAlpine, L. and Inouye, K. (2021). What value do PhD graduates offer? An organizational case study, Higher Education Research & Development, pp.1-16. ↩

  39. BEIS (2017). Industrial Strategy - Building a Britain fit for the future. London: BEIS (Department for Business Energy & Industrial Strategy). ↩

  40. Policy Links (2021). UK Innovation Report. Benchmarking the UK’s Industrial and Innovation Performance in a Global Context. Institute for Manufacturing - University of Cambridge ↩

  41. Smith, E. and White, P. (2018). Where Do All the STEM Graduates Go? Higher Education, the Labour Market and Career Trajectories in the UK. Journal of Science Education and Technology, 28(1), pp. 26-40. ↩

  42. Bove, V. and Elia, L. (2017). Migration, Diversity, and Economic Growth. World Development, 89, pp. 227-239. ↩

  43. Docquier, F., Turati, R., Valette, J. & Vasilakis, C. (2019). Birthplace diversity and economic growth: evidence from the US states in the Post-World War II period. Journal of Economic Geography, 20(2), pp. 321-354. ↩

  44. Trax, M., Brunow, S. & Suedekum, J. (2015). Cultural diversity and plant-level productivity. Regional Science and Urban Economics, 53, pp. 85-96. ↩

  45. Kemeny, T. and Cooke, A. (2018). Spillovers from immigrant diversity in cities. Journal of Economic Geography, 18(1), pp. 213-245. ↩

  46. Roupakias, S. and Dimou, S. (2019). Impact of cultural diversity on local labor markets. Evidence from Greece’s “age of mass migration”. The Manchester School, 88(2), pp. 282-304. ↩

  47. Ozgen, C., Nijkamp, P. & Poot, J. (2013). The impact of cultural diversity on firm innovation: evidence from Dutch micro-data. IZA Journal of Migration, 2(1), pp. 1. ↩

  48. Rodríguez-Pose, A. and Hardy, D. (2015). Cultural Diversity and Entrepreneurship in England and Wales. Environment and Planning A: Economy and Space, 47(2), pp. 392-411. ↩

  49. Østergaard, C. R., Timmermans, B. & Kristinsson, K. (2011). Does a different view create something new? The effect of employee diversity on innovation. Research Policy, 40(3), pp. 500-509. ↩

  50. Ritter-Hayashi, D., Vermeulen, P. & Knoben, J. (2019). Is this a man’s world? The effect of gender diversity and gender equality on firm innovativeness. PLoS One, 14(9), pp. e0222443. ↩

  51. British Science Association (2021). Inquiry into Equity in the STEM Workforce. All-Party Parliamentary Group (APPG) on Diversity and Inclusion in STEM. ↩

  52. White, P. and Smith, E. (2021). From subject choice to career path: Female STEM graduates in the UK labour market. Oxford Review of Education, pp. 1-17. ↩

  53. Britton, J., Dearden, L., Shephard, N. & Vignoles, A. (2016). How English domiciled graduate earnings vary with gender, institution attended, subject and socio-economic background. Institute for Fiscal Studies. IFS Working Paper W16/06. ↩

  54. Walker, I. and Zhu, Y. (2018). University selectivity and the relative returns to higher education: Evidence from the UK. Labour Economics, 53, pp. 230-249. ↩

  55. Zwysen, W. and Longhi, S. (2018). Employment and earning differences in the early career of ethnic minority British graduates: the importance of university career, parental background and area characteristics. Journal of Ethnic and Migration Studies, 44(1), pp. 154-172. ↩

  56. Schneider, P. and Sting, F. J. (2020). Employees’ Perspectives on Digitalization-Induced Change: Exploring Frames of Industry 4.0. Academy of Management Discoveries, 6(3), pp. 406-436. ↩

  57. Pagán-Rodríguez, R. (2014). Disability, Training and Job Satisfaction. Social Indicators Research, 122(3), pp. 865-885. ↩

  58. Böckerman, P. and Ilmakunnas, P. (2012). The Job Satisfaction-Productivity Nexus: A Study Using Matched Survey and Register Data. ILR Review, 65(2), pp. 244-262. ↩

  59. ONS, Employment in research and development occupations by nationality, selected years 2009 to 2023 ↩

  60. ONS, Employment in research occupations and development professions by ITL2 area, UK, 2004 to 2020 ↩

  61. Annual population survey (APS) QMI ↩

  62. Available at: Table 4 - HE academic staff by ethnicity and academic employment function 2014/15 to 2023/24. Note: In 2019/2020, additional providers were added to the coverage of the HESA Staff return (largely independent and private providers registered in England) and subsequently began reporting their Academic staff to HESA for the first time. Analysis of HESA returns data suggests this accounts for less than 1% of total staff numbers and likely has a minimal impact on the overall time series. ↩

  63. Some job titles associated with the SOC 2010 code “2139 – IT & telecommunications professionals n.e.c.” include IT consultants, software testers, telecommunications planners, and quality analysts. ↩

  64. Jerrim, J. and Schoon, I. (2014) Do teenagers want to become scientists?: A comparison of gender differences in attitudes toward science, career expectations, and academic skill across 29 countries. In: Schoon, I. and Eccles, J. S. (Eds.) Gender Differences in Aspirations and Attainment: A Life Course Perspective. Cambridge: Cambridge University Press, pp. 203-223. ↩

  65. Forth, T. and Jones, R. (2020). The Missing £4 Billion - Making R&D work for the whole UK. ↩

  66. See the relevant note of Table A5 in the accompanying Annex data tables regarding the countries that comprise each nationality group. ↩

  67. Includes specific learning difficulties, such as dyslexia or, dyscalculia, and severe learning difficulties, including mental impairments. ↩

  68. The small sample size of people with learning difficulties prevents us from reporting the relevant figures for each UK region/nation. ↩

  69. World Economic Forum (2020). The Future of Jobs Report 2020. ↩

  70. Enterprise Research Centre (2022). The State of Small Business Britain 2021: Enabling the Triple Transition. ↩

  71. ONS (2021a). Changing trends and recent shortages in the labour market, UK: 2016 to 2021. ONS (Office for National Statistics). ↩

  72. OECD (2021). An Assessment of the Impact of COVID-19 On Job and Skills Demand Using Online Job Vacancy Data. Paris: OECD Publishing. ↩

  73. Similar conclusions are drawn when looking at the “broad” definition of the R&D workforce (see Table A7 in the accompanying Annex data tables). ↩

  74. Green, F. (2016). Skills Demand, Training and Skills Mismatch: A Review of Key Concepts, Theory and Evidence. London: Government Office for Science. ↩

  75. Cardenas Rubio, J. and Hogarth, T. (2021). The R&D Pipeline. BEIS (Department for Business Energy & Industrial Strategy) - Warwick Institute of Employment Research. BEIS Research Paper Number: 2021/22.  ↩

  76. Gambin, L., Hogarth, T., Murphy, L., Spreadbury, K., Warhurst, C. & Winterbotham, M. (2016). Research to understand the extent, nature and impact of skills mismatches in the economy. BIS (Department for Business Innovation and Skills). Research Paper 265.  ↩

  77. Details about the independent variables included in each Model are provided in the notes of Table A10 in the accompanying Annex data tables and outlined in the Technical Annex of this report (Section 10.4).  ↩

  78. “Unobserved characteristics” are characteristics that we do not have measurements for in the data.  ↩

  79. Waldfogel, J. (1998). Understanding the ‘Family Gap’ in Pay for Women with Children. Journal of Economic Perspectives, 12(1), pp. 137-156.  ↩

  80. Viitanen, T. (2012). The motherhood wage gap in the UK over the life cycle. Review of Economics of the Household, 12(2), pp. 259-276.  ↩

  81. UKIS defines “innovation active” firms as firms who did any of the following activities during the survey period:
    1. The introduction of a new or improved product (goods or services);
    2. Business processes used to produce or supply all goods or services that the business has introduced, regardless of their origin. These innovations may be new to business or new to the market;
    3. Engagement in innovation projects not yet complete or abandoned;  ↩

  82. DBT (2024). UK Innovation Survey 2023. DBT (Department for Business and trade)  ↩

  83. LinkedIn - Industries with the Highest (and Lowest) Turnover Rates.  ↩

  84. See further details here at www.mitalent.ac.uk  ↩

  85. The TALENT Commission (2022). Technical skills, roles and careers in UK higher education and research.  ↩

  86. Taylor, A., Nelson, J., O’Donnell, S., Davies, E. & Hillary, J. (2022). The Skills Imperative 2035: what does the literature tell us about essential skills most needed for work? Slough: NFER  ↩

  87. Research and innovation (R&I) workforce survey report, 2022  ↩

  88. Belt, V., Ri, A. & Akinremi, T. (2021). The UK’s business R&D workforce: skills, sector trends and future challenges. ERC (Enterprise Research Centre).  ↩

  89. Ali, W., Bekiros, S., Hussain, N., Khan, S. A. & Nguyen, D. K. (2023). Determinants and consequences of corporate social responsibility disclosure: A survey of extant literature. Journal of Economic Surveys.  ↩

  90. ONS (2021b). Gross domestic expenditure on research and development, by region, UK. ONS (Office for National Statistics).  ↩

  91. See Tables A13 and A14 in the accompanying Annex data tables for more information.  ↩

  92. 4-digit SOC data available here: https://www.gov.uk/government/publications/labour-market-and-skills-projections-2020-to-2035  ↩

  93. ONS Number of People in Employment (aged 16 and over, seasonally adjusted)  ↩

  94. ONS, UK gross domestic expenditure on research and development (designated as official statistics)  ↩

  95. Bove, V. and Elia, L. (2017). Migration, Diversity, and Economic Growth. World Development, 89, pp. 227-239.  ↩

  96. Docquier, F., Turati, R., Valette, J. & Vasilakis, C. (2019). Birthplace diversity and economic growth: evidence from the US states in the Post-World War II period. Journal of Economic Geography, 20(2), pp. 321-354.  ↩

  97. Trax, M., Brunow, S. & Suedekum, J. (2015). Cultural diversity and plant-level productivity. Regional Science and Urban Economics, 53, pp. 85-96.  ↩

  98. Kemeny, T. and Cooke, A. (2018). Spillovers from immigrant diversity in cities. Journal of Economic Geography, 18(1), pp. 213-245.  ↩

  99. Roupakias, S. and Dimou, S. (2019). Impact of cultural diversity on local labor markets. Evidence from Greece’s “age of mass migration”. The Manchester School, 88(2), pp. 282-304.  ↩

  100. Ozgen, C., Nijkamp, P. & Poot, J. (2013). The impact of cultural diversity on firm innovation: evidence from Dutch micro-data. IZA Journal of Migration, 2(1), pp. 1.  ↩

  101. Østergaard, C. R., Timmermans, B. & Kristinsson, K. (2011). Does a different view create something new? The effect of employee diversity on innovation. Research Policy, 40(3), pp. 500-509.  ↩

  102. Ritter-Hayashi, D., Vermeulen, P. & Knoben, J. (2019). Is this a man’s world? The effect of gender diversity and gender equality on firm innovativeness. PLoS One, 14(9), pp. e0222443.  ↩

  103. Rodríguez-Pose, A. and Hardy, D. (2015). Cultural Diversity and Entrepreneurship in England and Wales. Environment and Planning A: Economy and Space, 47(2), pp. 392-411.  ↩

  104. Belt, V., Ri, A. & Akinremi, T. (2021). The UK’s business R&D workforce: skills, sector trends and future challenges. ERC (Enterprise Research Centre).  ↩

  105. OECD (2015). Frascati Manual 2015 - Guidelines for Collecting and Reporting Data on Research and Experimental Development Paris: OECD Publishing. Available from https://www.oecd.org/publications/frascati-manual-2015-9789264239012-en.htm.  ↩

  106. See SOC 2010 volume 1: structure and descriptions of unit groups for the SOC 2010 hierarchy, job descriptions and tasks required by each occupation. Similarly, see SOC 2020 Volume 1: structure and descriptions of unit groups for the corresponding information for the SOC 2020. The codification changes because of the transition from SOC 2010 to SOC 2020 affect only the year 2021 in our data.  ↩

  107. The TALENT Commission (2022). Technical skills, roles and careers in UK higher education and research.  ↩

  108. Belt, V., Ri, A. & Akinremi, T. (2021). The UK’s business R&D workforce: skills, sector trends and future challenges. ERC (Enterprise Research Centre).  ↩

  109. Employment in research and development occupations by nationality, selected years 2009 to 2023 and Employment in research occupations and development professions by ITL2 area, UK, 2004 to 2020  ↩

  110. The Annual population survey (APS) datasets are weighted by the ONS to reflect the size and composition of the general population using the most up-to-date official population data. For more information see the Annual population survey (APS) QMI  ↩

  111. Office for National Statistics, Social Survey Division. (2022). Annual Population Survey, 2004-2021: Secure Access. [data collection]. 22nd Edition. UK Data Service. SN: 6721, DOI: http://doi.org/10.5255/UKDA-SN-6721-22. Because of the sensitive nature of the data, we conducted the analysis remotely within the Secure Lab of the UK Data Service. Based on the Statistical Disclosure Control, the underlying sample size of each cell of the reported tables/graphs should comprise at least 10 people to publish the results.  ↩

  112. Note that the impact of the SOC coding issue identified by the ONS for 2021 data is expected to be minimal on our analysis, which predominantly utilises pooled APS data from July 2012 to June 2021. For detailed information on the revision of miscoded occupational data in the Labour Force Survey, refer to the ONS statement.  ↩

  113. Blau, P. M., Gustad, J. W., Jessor, R., Parnes, H. S. & Wilcock, R. C. (1956). Occupational Choice - a Conceptual Framework. Industrial & Labor Relations Review, 9(4), pp. 531-543.  ↩

  114. Dolton, P. J., Makepeace, G. H. & Vanderklaauw, W. (1989). Occupational Choice and Earnings Determination - the Role of Sample Selection and Non-Pecuniary Factors. Oxford Economic Papers, 41(3), pp. 573-594.  ↩

  115. Drost, A. (2002). The Dynamics of Occupational Choice: Theory and Evidence. Labour, 16(2), pp. 201-233.  ↩

  116. Xu, Y. J. (2012). Career Outcomes of STEM and Non-STEM College Graduates: Persistence in Majored-Field and Influential Factors in Career Choices. Research in Higher Education, 54(3), pp. 349-382.  ↩

  117. Holmes, K., Gore, J., Smith, M. & Lloyd, A. (2017). An Integrated Analysis of School Students’ Aspirations for STEM Careers: Which Student and School Factors Are Most Predictive? International Journal of Science and Mathematics Education, 16(4), pp. 655-675.  ↩

  118. Speer, J. D. (2017). Pre-Market Skills, Occupational Choice, and Career Progression. Journal of Human Resources, 52(1), pp. 187-246.  ↩

  119. Lent, R. W., Sheu, H. B., Miller, M. J., Cusick, M. E., Penn, L. T. & Truong, N. N. (2018). Predictors of science, technology, engineering, and mathematics choice options: A meta-analytic path analysis of the social-cognitive choice model by gender and race/ethnicity. Journal of Counseling Psychology, 65(1), pp. 17-35.  ↩

  120. Cardenas Rubio, J. A. (2020). A web-based approach to measure skill mismatches and skills profiles for a developing country: the case of Colombia. Bogotá: Editorial Universidad del Rosario.  ↩

  121. MAC (2017). Assessing labour market shortages: A methodology update. London: Migration Advisory Committee.  ↩

  122. UK Commission for Employment and Skills, Department for Education. (2023). Employer Skills Survey, 2011-2019: Secure Access. [data collection]. 6th Edition. UK Data Service. SN: 7745, DOI: http://doi.org/10.5255/UKDA-SN-7745-6  ↩

  123. Galindo-Rueda, F. and Verger, F. (2016). OECD taxonomy of economic activities based on R&D intensity. OECD Publishing. No. 2016/04.  ↩

  124. Carvalho, A. (2022). R&D intensity at industry level: how does UK compare with top performing OECD countries? Enterprise Research Centre.  ↩

  125. Winterbotham, M., Kik, G., Selner, S., Menys, R., Stroud, S. & Whittaker, S. (2020). Employer Skills Survey 2019: Research report. DfE (Department for Education). DFERPPU 2018061/2.  ↩

  126. UK gross domestic expenditure on research and development  ↩

  127. WMCA (2021). Supplementary written evidence submitted by the West Midlands Combined Authority (CRV0069). Written submission following the oral evidence from Andy Street, Mayor of the West Midlands, Dr Clive Hiickman, Chief Executive of the Manufacturing Technology Centre and Prof Simon Collinson, Director of WMREDI and Professor of International Business & Innovation on 25 November 2021.  ↩

  128. See The Skills Imperative 2035 - NFER  ↩

  129. Wilson, R., Hillary, J., Bosworth, D., Bosworth, L., Cardenas-Rubio, J., Day, R., Patel, S., Bui, H., Lin, X., Seymour, D. & Thoung, C. (2022c). The Skills Imperative 2035: Occupational Outlook – Long-run employment prospects for the UK. Headline Report. Slough: NFER.  ↩

  130. Wilson, R., Bosworth, D., Bosworth, L., Cardenas-Rubio, J., Day, R., Patel, S., Bui, H., Lin, X., Seymour, D. & Thoung, C. (2022b). The Skills Imperative 2035: Occupational Outlook – Long-run employment prospects for the UK, Baseline Projections. Working Paper 2a. Slough: NFER.  ↩

  131. Wilson, R., Bosworth, D., Bosworth, L., Cardenas-Rubio, J., Day, R., Patel, S., Bui, H., Lin, X., Seymour, D. & Thoung, C. (2022a). The Skills Imperative 2035: Occupational Outlook – Long-run employment prospects for the UK, Alternative scenarios. Working Paper 2b. Slough: NFER.  ↩

  132. Data available via GOV.UK Education Statistics: https://www.gov.uk/government/publications/labour-market-and-skills-projections-2020-to-2035  ↩

  133. Unit for Future Skills Jobs and Skills Dashboard  ↩

  134. Green, A. (2016) Low skill traps in sectors and geographies: underlying factors and means of escape, Foresight, Government Office for Science.  ↩

  135. Green, F. and Ashton, D. (1992) Skill shortage and skill deficiency: a critique, Work, Employment & Society 6(2), pp. 287-301.  ↩

  136. Oliver, J.M. and Turton, J.R. (1982) Is there a shortage of skilled labour?, British Journal of Industrial Relations 20(2), pp. 195-200.  ↩

  137. Gambin, L., Hogarth, T., Murphy, L., Spreadbury, K., Warhurst, C. & Winterbotham, M. (2016). Research to understand the extent, nature and impact of skills mismatches in the economy. BIS (Department for Business Innovation and Skills). Research Paper 265.  ↩

  138. MAC (2017). Assessing labour market shortages: A methodology update. London: Migration Advisory Committee.  ↩

  139. Allen, J. and van der Velden, R. (2001) Educational mismatches versus skill mismatches: effects on wages, job satisfaction, and on-the-job search, Oxford Economic Papers 53(3), pp. 434- 452.  ↩

  140. McGuinness, S. and Sloane, P.J. (2011) Labour market mismatch among UK graduates: An analysis using REFLEX data, Economics of Education Review 30(1), pp. 130-145.  ↩

  141. Kampelmann, S. and Rycx, F. (2012) The impact of educational mismatch on firm productivity: Evidence from linked panel data, Economics of Education Review 31(6), pp. 918-931.  ↩

  142. Green, F. and Henseke, G. (2021) Task-warranted graduate jobs and mismatch, Singapore Economic Review.  ↩

  143. Lyons, H., Taylor, A. and Green, A. (2020) Rising to the UK’s Skills Challenges, Industrial Strategy Council.  ↩

  144. Galindo-Rueda, F. and Verger, F. (2016). OECD taxonomy of economic activities based on R&D intensity. OECD Publishing. No. 2016/04.  ↩

  145. Carvalho, A. (2022). R&D intensity at industry level: how does UK compare with top performing OECD countries? Enterprise Research Centre.  ↩

  146. Marra, A. and Baldassari, C. (2022). Using text data instead of SIC codes to tag innovative firms and classify industrial activities. PloS One, 17(6), pp. e0270041-e0270041.  ↩

  147. OECD (2021) Burning Glass Technologies’ data use in policy-relevant analysis: An occupation level assessment  ↩

  148. Mood, C. (2009). Logistic Regression: Why We Cannot Do What We Think We Can Do, and What We Can Do About It. European Sociological Review, 26(1), pp. 67-82.  ↩

  149. Amemiya, T. (1981). Qualitative Response Models: A Survey. Journal of Economic Literature, 19(4), pp. 1483.  ↩

  150. Because of the lack of information on workers’ experience, a typical method adopted in empirical studies is to use age as a proxy for experience. In addition, we include the term “age squared” in the regression model to capture the likely diminishing impact of age on the dependent variable (i.e., the inverted U-shaped relationship). Specifically, the probability of conducting R&D activities should increase with age until a particular high point (for example, until the early 40s) and likely decrease after that. Moreover, the survey year dummies control for the fact that the probability of being an R&D worker is different across the years of our study.  ↩

  151. Long, J. S. and Freese, J. (2014) Regression models for categorical dependent variables using Stata.  Texas: Stata Press.  ↩

  152. Williams, R. (2012). Using the margins command to estimate and interpret adjusted predictions and marginal effects. Stata Journal, 12(2), pp. 308-331.  ↩

  153. Schoon, I. and Parsons, S. (2002). Teenage Aspirations for Future Careers and Occupational Outcomes. Journal of Vocational Behavior, 60(2), pp. 262-288.  ↩

  154. Drost, A. (2002). The Dynamics of Occupational Choice: Theory and Evidence. Labour, 16(2), pp. 201-233.  ↩

  155. Source: What qualification levels mean  ↩

Back to top