Skill relatedness: A dynamic approach to regional diversification in the UK

Abstract

Exploring the factors driving new industrial activities is vital to regional economies. Regional diversification typically involves leveraging local capabilities related to those underpinning their existing industrial base to develop new industrial specialisations that branch out from established ones, which is captured by the concept of relatedness. Despite growing scholarly interest, our understanding of how relatedness drives regional diversification across varied regional contexts remains limited. This study addresses this gap by focusing on skills as the micro-foundations of regional capabilities and examining how skill relatedness across two distinct types: technical-digital (TD) and social-cognitive (SC), shapes the likelihood of new manufacturing specialisations across four types of UK regions. Our findings show that the marginal effect of TD skill relatedness on regional diversification is stronger in left-behind regions and that SC skill relatedness further amplifies this effect. However, the magnitude and nature of these effects vary across regional contexts. These results enrich evolutionary economic geography debates by embedding insights into how different skill compositions and their interactions shape the place-based dynamics of regional diversification. We conclude by discussing policy implications for integrating regional skill policies within place-based industrial strategies and identifying avenues for future research.

Keywords

Introduction

Skills are widely recognised as fundamental drivers of regional productivity and economic growth (Moretti, 2012; Plummer and Taylor, 2001). A strong base of skilled labour forms the micro-foundations of regional capabilities (Whittle and Kogler, 2020), enabling cross-industry labour mobility as workers transition between industries requiring similar skills (Neffke and Henning, 2012). Skills thus create opportunities for knowledge recombination and the emergence of new industrial specialisations (Corradini et al., 2022), a process commonly referred to as regional industrial diversification (Boschma, 2016; Deegan et al., 2024).

Increasingly, this process is emphasised within Evolutionary Economic Geography (EEG) through the conceptual lens of skill relatedness, which posits regions are more likely to develop new industrial specialisations by building upon their existing skills (Fitjar and Timmermans, 2016; Henning et al., 2025; Neffke and Henning, 2012;Whittle and Kogler, 2020). Yet, current EEG contributions face three key limitations. First, while skills are often put at the forefront of analyses on regional development, regional skills remain under-theorised and empirically challenging to capture (Henning et al., 2025). The existing literature largely operationalises skill relatedness indirectly through cross-industry labour flows, which may lead to overestimation, as labour mobility can also reflect corporate culture and network effects (Buyukyazici et al., 2023; Neffke and Henning, 2012). Second, skills are often operationalised as a single, aggregated construct, despite increasing evidence that technological change is fundamentally reshaping the composition of skills required across industries (Ciarli et al., 2021; De Propris and Bellandi, 2021; Poláková et al., 2023). As a result, we know relatively little about how specific types of skills and their interactions affect regional diversification (Buyukyazici et al., 2023; Castellacci et al., 2019; Santoalha et al., 2021). Third, traditional research on regional industrial diversification often adopts a place-neutral approach, offering limited insight into how skill relatedness leads to varied diversification outcomes across different regional contexts (Boschma, 2016). Despite growing recognition that industrial diversification is fundamentally a place-based process (Kogler et al., 2023), empirical studies often rely on simplified distinctions defined solely by development levels (e.g. core/peripheral, leading/lagging regions) (Frenken et al., 2007; Pinheiro et al., 2022). Yet, beyond development levels, regions differ along a range of historically embedded structural features such as institutional configurations, industrial networks, and historical trajectories, which influence both the composition of the local skill base and the regional capacity to recombine these skills in support of new industrial activities (Bathelt and Storper, 2023). As a result, the extent to which regions can exploit skill relatedness across different domains for diversification varies significantly. Advancing theoretical understanding within EEG thus calls for a more fine-grained geographical perspective that systematically accounts for these broader institutional and contextual elements (Bathelt and Storper, 2023; Boschma, 2016; Xiao et al., 2018).

This article addresses these key gaps by utilising skill demand information derived from 955,821 Lightcast UK job postings, spanning 166 British International Territorial Level (ITL)-3 level regions,¹ to assess the skill demand portfolio of regions and measure skills relatedness at the industry-region level. We empirically examine how four types of UK regions (characterised by distinct combinations of development level and historical industrial heritage) diversify their manufacturing sectors by leveraging skill relatedness across two distinct domains: technical-digital and social-cognitive (SC) skills. Our analysis seeks to understand (i) how skill relatedness within these two distinct domains, and their interaction, shape the likelihood of new regional industrial specialisations in manufacturing industries; and (ii) how the impact of skill relatedness on industrial diversification patterns varies across the four regional contexts.

This study makes three key contributions. First, it advances the EEG literature on skill relatedness by bridging two previously disconnected strands of inquiry: the evolving impact of technological change on the composition of skill demand and the place-based dynamics of regional diversification. We provide new insights into how different regions diversify their manufacturing sectors by signalling demand for distinct combinations of related skills across two domains in the context of technological change, thereby broadening the theoretical scope of the skill-relatedness framework. Second, the study links theoretical analysis with practical policy development, offering insights for place-based industrial strategies. It calls for a shift away from conventional reactive skills policies and emphasises the need to integrate skills policy more proactively into industrial strategy to better support regional diversification (Corradini et al., 2022). Third, the study offers several methodological contributions: it develops a forward-looking approach for measuring skill relatedness based on skill demand information derived from detailed job postings; applies text-mining techniques to systematically classify the full skill set into two distinct domains; and develops a novel 2 × 2 regional typology that captures regional variation in both development level and industrial heritage.

The remainder of the article is as follows. “Literature review and research hypotheses” section presents our theoretical framework and research hypotheses. “Data and methodology” section outlines our data and model specification(s). “Results and discussion” section discusses our results. Finally, “Conclusion” section considers policy implications and concludes.

Literature review and research hypotheses

Technological change, evolving skill demands, and skill relatedness in regional diversification: Differentiating two skill domains

A skill is defined as the ability to apply knowledge and use know-how to complete tasks and solve problems² (Henning et al., 2025). At the job level, they represent a fundamental unit underpinning several closely related concepts, such as knowledge, abilities, and tasks (Henning et al., 2025). At the regional level, they constitute the micro-foundations of regional capabilities necessary to perform tasks, enable learning processes, generate competitive advantage, and foster new specialisations (Plummer and Taylor, 2001). A strong base of skilled labour is widely recognised as a key driver of local productivity, innovation, and competitiveness (Morris et al., 2019).

The centrality of skills in underpinning regional competitiveness is particularly relevant when considering the implementation of regional innovation strategies (Clark and Bailey, 2018). Recent policy and academic debates point to regional development being built upon place-based capabilities (Feldman and Kogler, 2010; McCann, 2019; von Hippel, 1994), often evolving incrementally along existing industrial trajectories (Castaldi and Drivas, 2023; Corradini and Vanino, 2022; Martin and Sunley, 2006). In this context, the effective utilisation and recombination of existing skills within a region have emerged as a viable strategy for regional diversification, facilitating the emergence of new industrial specialisations that branch out from established ones (Corradini et al., 2022).

A growing body of research within EEG has conceptualised this process through the lens of skill relatedness, defined as the degree of similarity in skill requirements between industries at the regional level (Fitjar and Timmermans, 2016; Galetti et al., 2021). In this framework, industries are considered more or less related based on the extent to which they draw upon similar skill sets embodied within the local workforce (Boschma et al., 2014; Neffke et al., 2017). A higher level of skill relatedness indicates that a new industry shares similar skill profiles to the industries already present in the region, thereby increasing the potential for new industrial specialisations to emerge by leveraging related skills embedded in the local workforce (Boschma et al., 2014; Corradini and Vanino, 2022; Neffke et al., 2017). However, research on skill relatedness remains at a relatively early stage. While EEG studies acknowledge the importance of diverse knowledge bases for regional productivity (Asheim et al., 2017; Corradini and Karoglou, 2022), the practical skillsets required to leverage them have largely been overlooked. Most EEG contributions operationalise skills in an aggregated form, with only limited work examining how skill relatedness within specific skill domains shapes regional diversification (Buyukyazici et al., 2023; Castellacci et al., 2019; Santoalha et al., 2021). However, ongoing rapid technological change, including automation, data integration, and the adoption of advanced technologies (Ciarli et al., 2021; De Propris and Bellandi, 2021), has been fundamentally reshaping task contents (Acemoglu and Restrepo, 2018), and, in turn, altering the demand for evolving skill compositions across sectors. This makes it essential to consider skill relatedness within specific skill domains.

Of particular relevance is the manufacturing sector, which has been undergoing a shift towards smart, digitally integrated production systems, driven by the fourth and fifth industrial revolutions (Frey and Osborne, 2017; Sunley et al., 2021). Within this context, technical-digital (TD) skills have received considerable emphasis in both academic research (Buyukyazici, 2024; Falck et al., 2021; Santoalha et al., 2021) and policy discourses (as reflected in EU Smart Specialisation Strategies; Hassink and Gong, 2019; Morgan and Marques, 2019). These skills typically include proficiency in operating digital technologies and specialised equipment, using software applications, programming, and data analytics (Castellacci et al., 2019; Ciarli et al., 2021). TD skills are particularly valuable for supporting codifiable, automatable, and engineering-based tasks, facilitating the processing of large datasets (Ciarli et al., 2021; Cirillo et al., 2021), and underpinning the adoption of advanced digital tools into tasks (Ciarli et al., 2021; De Propris and Bellandi, 2021).

However, emerging visions of Industry 5.0 increasingly emphasise the complementary role of SC skills (Poláková et al., 2023), which encompass capabilities related to interpersonal collaboration and communication, informal networking, innovative and critical thinking, and adaptive problem-solving (Alan et al., 2019; Deming and Kahn, 2018; Whittemore, 2018). These human-centric capabilities, which remain difficult to automate (Autor et al., 2015; UKCES, 2015), are becoming increasingly essential for enabling the practical, responsible, and intelligent use of new technologies (Acemoglu and Restrepo, 2019; Falck et al., 2021). As such, SC skills may play a crucial complementary role by enabling TD skills to realise their full potential in driving technological progress.

For regions seeking to leverage skill relatedness to develop new manufacturing specialisations, it is critical to account for the alignment of skills across both domains. This study distinguishes between TD and SC skill relatedness, reflecting the evolving composition of skills demand in manufacturing in the context of ongoing technological change. Given the growing importance of SC skills in supporting the adoption and effective application of new technologies (Poláková et al., 2023), we anticipate a complementary interaction whereby SC skill relatedness strengthens the effect of TD skill relatedness on regional diversification. Hence, we posit:

H1: The interaction between TD and SC skill relatedness yields higher probabilities of regional diversification into new manufacturing industries, beyond the effect of TD skill relatedness alone.

Place and path dependencies: Varied regional diversification outcomes

So far, our discussion has adopted a place-neutral perspective, reflecting another common limitation in existing EEG contributions, which often generalise the role of skill relatedness across regions (Boschma, 2016). However, it is important to note that regions may vary significantly in their capacity to recombine existing skills to support the development of new industrial specialisations; as a result, skill relatedness may lead to distinct diversification outcomes. This variation can be understood along two dimensions.

First, the place-specific nature of labour market contexts, evident in the varied demand and supply schedules for different skillsets across regions (Corradini et al., 2022), may result in variable regional capacities to utilise related skills across different skill domains. Given the persistence and widening of digital divides³ (Reveiu et al., 2022; Salemink et al., 2017), such regional variation is likely to be especially pronounced within the domain of TD skills, particularly among regions with different development levels (Eriksson and Hane-Weijman, 2016; McCann, 2019).

More advanced regions are typically positioned closer to the technological frontier, underpinned by a stronger base of TD skills needed to adopt, develop, and leverage emerging technologies. These regions tend to benefit from a tighter alignment between skill supply and industrial demand (Bacolod et al., 2009; Glaeser et al., 2012), placing them in a favourable position for leveraging related TD skills to support knowledge-intensive and technologically sophisticated manufacturing activities. However, as many of these regions are already highly diversified, further increases in TD skill relatedness may generate only limited additional benefits due to diminishing marginal returns (Katz and Shapiro, 1985). Moreover, TD skills in these regions are more commonly oriented towards the development of advanced or frontier technologies, which typically entail longer time horizons and substantial resource commitments before they can be translated into viable new specialisations, particularly in manufacturing industries with higher technological thresholds. As a result, the short- to medium-term effects of improved TD skill relatedness on diversification may be relatively attenuated in these regions.

In contrast, left-behind regions, characterised by low development, slow innovation, stagnating employment, and weaker social dynamics (Eder and Trippl, 2019; Mayer, 2020; Rodríguez-Pose et al., 2023; Velthuis et al., 2024), face enduring barriers, such as distance from the technological frontier, limited digital infrastructure and a workforce largely possessing only basic ICT skills, often confining them to low-tech manufacturing. However, starting from a lower base, these regions may experience stronger marginal gains from improvements in TD skill relatedness. Such improvements can facilitate the adoption of already-established technologies new to the region, laying the groundwork for new industrial specialisations, particularly in industries with lower technological thresholds (Barzotto et al., 2020). This process often occurs through imitation and adaptation of successful practices from more advanced regions (Grillitsch et al., 2018; Isaksen et al., 2018). Consequently, the same increase in TD skill relatedness may yield stronger diversification outcomes in left-behind regions than in more advanced ones, where such skills may already be prevalent. Hence, we posit:

H2: The marginal effect of TD skill relatedness on regional diversification is stronger in left-behind regions.

Second, consistent with the path-dependent nature of regional diversification processes, new specialisations tend to emerge incrementally along existing industrial trajectories (Castaldi and Drivas, 2023; Corradini and Vanino, 2022; Martin and Sunley, 2006). Institutional and contextual factors, such as prevailing industrial structures, entrenched networks, and inherited institutional configurations, also shape a region’s capacity to recombine existing skills and foster new industrial specialisations branching out of existing ones (Bathelt and Storper, 2023).

In some cases, these structures, while effective in the past, may become so entrenched and inflexible that they no longer align with the demands of a changing industrial and technological environment (Roessler, 2025). This is particularly relevant for some old industrial regions, where deeply embedded manufacturing traditions, production structures, and prevailing institutions may constrain the scope for diversification by locking regions into rigid historical trajectories, reinforcing patterns of stagnation and decline (Hassink, 2010; Sunley and Martin, 2023; Tödtling and Trippl, 2013).

Thus, it is important to consider these institutional dynamics alongside varying levels of regional development, as they may jointly influence a region’s capacity to leverage related skill across TD and SC domains. In response, we move beyond typologies based solely on development levels and incorporate industrial heritage as an additional axis of differentiation. Accordingly, we posit:

H3: The effects of TD and SC skill relatedness (and their interaction) on regional diversification vary across regions with different development levels and industrial heritage.

Data and methodology

Data sources

This study utilises secondary data from two main sources. First, we join an emerging strand of research that examines the evolution and spatial patterns of skill demands expressed in job posting data (Acemoglu et al., 2022; Deming and Noray, 2020; Garcia-Lazaro et al., 2025; Henning et al., 2025), which may help avoid the overestimation of skills associated with traditional indirect measures based on cross-industry labour flows (Buyukyazici et al., 2023; Neffke and Henning, 2012). Our analysis is based on 955,821 unique Lightcast job postings from firms in manufacturing industries⁴ between 2014 and 2022. Lightcast is a labour market analytics platform that aggregates job advertisements from multiple sources, including employer websites, applicant tracking systems, and online job boards. Each posting contains detailed information on job title, description, required skills, wage, location, occupation, and industry classification.

Second, we draw on socioeconomic indicators from the Office for National Statistics (ONS) to categorise regions, construct control variables, and the dependent variables Entry and Exit.

Construction of TD and SC skill typology

To classify the whole skillset into two mutually exclusive clusters: TD skills and SC skills, we build on existing research (Alabdulkareem et al., 2018; Castellacci et al., 2019; Deming and Kahn, 2018; Rayna and Striukova, 2021) and create a compendium of relevant keywords and phrases for each cluster based on their significance and frequency of occurrence within the texts of each category (see Supplemental Appendix A.1 for the detailed compendium).

Utilising the compendium, we employ a Dictionary-Based Text Classification Method (Mandal and Gupta, 2016; Reveilhac and Morselli, 2022) to scan all the listed skills requirements in each UK online job posting for the presence of keywords from the TD and SC categories. Where classification confidence is low or ambiguous, results are manually reviewed and adjusted.

Construction of regional typology

We construct a regional typology by jointly considering regions’ development levels and industrial heritage. First, for the development level dimension, we classify regions as left-behind and non-left-behind based on their developmental trajectories between 2014 and 2022. This classification is derived using sequence analysis on a set of socioeconomic indicators, applying a combination of k-means clustering and optimal matching (Le Petit-Guerin et al., 2023; Velthuis et al., 2024). Left-behind regions are characterised by persistently below-average GDP per capita, low or negative employment growth, and stagnant or declining population levels (Barzotto et al., 2020; Rodríguez-Pose et al., 2023; Velthuis et al., 2024). Second, to capture the dimension of industrial heritage, we draw on historical employment data from 1971 to identify regions with a strong industrial heritage. Following Sunley et al. (2021), we classify a region as an old industrial region if the combined employment in manufacturing and mining exceeded 1 standard deviation (SD) above the national mean (i.e. above 33.8% of total employment).

Combining these two dimensions yields a 2 × 2 regional typology: Legacy Encumbered Regions (LERs), Legacy Void Regions (LVRs), Legacy Leveraged Regions (LLRs), and Modern Dynamic Regions (MDRs; see Supplemental Appendix A.2 for details).

LERs and LVRs represent two types of left-behind regions differentiated by their industrial heritage. LERs are left-behind old industrial regions characterised by unsuccessful industrial restructuring, where entrenched institutional structures and production networks encumber adaptation. Deindustrialisation has further exacerbated their decline (Hassink, 2010). In contrast, LVRs are left-behind regions that lack a manufacturing legacy and experience persistent economic stagnation; they are often disconnected from major growth dynamics (MacKinnon et al., 2022; Rodríguez-Pose, 2018).

LLRs and MDRs represent two types of non-left-behind regions, also differentiated by their industrial heritage. LLRs are old industrial regions that have successfully reconfigured and leveraged their deep-rooted industrial legacy to support diversification and renewal (Tierney et al., 2023). Whereas MDRs are among the economically strongest, encompassing the UK’s golden triangle of Oxford, London, and Cambridge. These regions were never historically dominated by a single industry (Sunley et al., 2021), yet they exhibit sustained economic strength, underpinned by innovation, advanced technological adoption, and dynamic urban labour markets (Fitjar and Rodríguez-Pose, 2013), positioning them on favourable development trajectories.

Dependent variable: Entry and Exit

Following the standard empirical practice in the literature that assesses regional diversification by tracking the emergence of new specialisations and/or the loss of existing (non-competitive) ones (Boschma et al., 2023; Buyukyazici et al., 2023; Kogler et al., 2023), we construct two binary dependent variables, Entry and Exit to capture the emergence and loss of regional industrial specialisations in manufacturing (classified using two-digit Standard Industrial Classification (SIC) codes 10–33 under Section C). We adopt standard practice in the diversification literature (Castellacci et al., 2019; Kogler et al., 2023), and operationalise regional industrial specialisation using the Revealed Comparative Advantage (RCA) index. A region is considered to specialise in a given industry if it achieves an RCA score greater than 1. Accordingly, Entry is coded as 1 when industry i in region p achieves RCA > 1 in period t but did not in t − 1,⁵ and 0 otherwise. Conversely, Exit is set to 1 if industry i had RCA > 1 in t − 1 but falls below this threshold in t (Hidalgo et al., 2007; Kogler et al., 2023; Neffke and Henning, 2012). More precisely:

E n t r y_{i, p, t} = 1, i f {RCA}_{i, p, t} > 1 a n d R C A_{i, p, t - 1} \leq 1

(1)

E x i t_{i, p, t} = 1, i f R C A_{i, p, t} \leq 1 a n d R C A_{i, p, t - 1} > 1

(2)

R C A_{i, p, t} = \frac{E_{i, p, t} / E_{i^{'}, p, t}}{E_{i, p^{'}, t} / E_{i^{'}, p^{'}, t}}

(3)

In these formulas, E_i,p,t is the employment in industry i in region p in period t. A higher RCA suggests that industry i possesses a greater employment share within region p at time t than its average employment across all regions p at the same time. In essence, an increased RCA suggests the industry’s strong presence in a specific region (Hidalgo et al., 2007).

Independent variable: Skill Relatedness Density

We build on Henning et al. (2025) and utilise expressed skill demands derived from Lightcast UK job postings to compute the independent variables of interest through a series of steps. Each Lightcast job posting is associated with a set of standardised skill identifiers, extracted and classified by Lightcast using proprietary algorithms applied to job descriptions (Tsvetkova et al., 2024). These identifiers capture specific abilities required to perform tasks and solve problems associated with the job. We first aggregate job postings at the industry-region-year level to construct industry-skill input matrices C_is for each region and year, where each cell (x_{i, s}) captures the intensity⁶ of skill s in industry i. Following Buyukyazici et al. (2023), we then utilise the Relative Skill Advantage (RSA) to assess the importance of a skill to a specific industry relative to its importance across all industries. A skill is considered to be relatively more strongly demanded by an industry if its RSA exceeds 1. RSA is defined at the skill-industry-region level, calculated as the ratio of the relative importance of skill s to industry i compared to the relative importance of the same skill across all industries in region p at time t. Formally:

RSA (i, s, t) = \frac{X_{i, s, t - 1} \ Σ_{s^{'} \in S} X_{i, s^{'}, t - 1}}{Σ_{i^{'} I} X_{i^{'} s, t - 1} \ Σ_{i^{'} \in I, s^{'} \in S} X_{i^{'}, s^{'}, t - 1}}

(4)

Where X_i,s,t_− 1 represents the demand of industry i for skill s in year t − 1 $Σ_{s^{'} \in S} X_{i, s^{'}, t - 1}$ is the total demand of industry i for all skills in year t − ; $Σ_{i^{'} \in I} X_{i^{'}, s, t - 1}$ is the total demand of all industries for skill s in year t − 1 and $Σ_{i^{'} \in I, s^{'} \in S} X_{i^{'}, s^{'}, t - 1}$ is the total demand of all industries for all skills in year t − 1.

Subsequently, we calculate the relatedness index within a network-based skill space, where two skills are connected based on the degree of their relatedness, measured by the minimum conditional probability of their co-occurrences across industry classes as formulated in equation (5):

R (s, s^{'}) = \frac{Σ_{i \in I} e (i, s) . e (i, s^{'})}{\max (Σ_{i \in I} e (i, s), Σ_{i \in I} e (i, s^{'})}

(5)

where relatively more strongly demanded skills denoted as e(i, s) = 1 if RSA > 1, and e(i, s) = 0 otherwise. The resulting skill relatedness index is an adjacency matrix of skills based on the co-occurrence analysis of industries in region p in year t. Applying equations (4) and (5) to all input matrices leaves a number of skill-to-skill relatedness indexes.

We combine the RSA with the co-occurrence matrix to derive a density indicator at the industry-region level, as in Boschma et al. (2014) and Buyukyazici et al. (2023):

Skill Relatedness {Density}_{i, t}^{p} = \frac{Σ_{s \in i} {(\frac{Σ_{s} \emptyset_{s, j, t} {RSA}_{s, i, t}}{Σ_{s} \emptyset_{s, j, t}} \times 100)}_{s, i, t}}{Σ_{s \in i} s}

(6)

Where $\emptyset_{s, j, t}$ refers to the relatedness between skill s and j at time t. ${RSA}_{s, i, t}$ is a binary variable that takes the value of 1 if skill s is relatively more strongly demanded by industry i at time t, and takes the value of 0 otherwise.

First, we calculate the skill-industry relatedness density matrix by dividing the sum of relatedness values between all skill pairs that industry i strongly demands by the sum of relatedness values across all skill pairs demanded in region p at time t (inside the parentheses in the numerator). This results in a relatedness density matrix between skill s and industry i, with each cell representing a skill relatedness density at the skill and industry level. We then aggregate the skill-industry level relatedness values to compute the industry-region Skill Relatedness Density (SRD) indicator. This is calculated as the sum of relatedness densities of skills that industry i strongly demands, divided by the sum of all skill relatedness densities demanded in region p at time t.

SRD is a forward-looking measure capturing the extent to which the skill space of a prospective industry resembles the contemporaneous regional skill demand structure, as reflected in job postings.⁷ A higher SRD indicates that the skills intensively demanded by industry i are more closely related to those currently demanded within region p. To examine the varied impacts of skill relatedness within the two domains, we compute SRD separately for TD and SC skills, resulting in two independent variables of interest: TD SRD and SC SRD. Each variable reflects how closely a potential new industry aligns with the region’s existing industry mix in terms of human capital requirements, within the domains of TD and SC skills, respectively.

Control variables

We incorporate an array of control variables across four categories. The first category includes the logarithm of Gross Domestic Expenditure on Research and Development (GERD), an indicator of regional investment in R&D fundamental for regional growth/innovation (Rodríguez-Pose and Crescenzi, 2008). The second includes the population density to capture regional urbanisation levels, which may affect local human capital and industrial transformation (Glaeser et al., 2012; Salvati and Carlucci, 2015). The third considers industrial-regional factors, capturing the dynamics within manufacturing through Industry Growth/Death, tracking birth and death rates of manufacturing industries in a region (Keeble and Walker, 2006). The fourth category captures geographical diversification, featuring metrics such as Industrial Ubiquity and Regional Diversity (Buyukyazici et al., 2023; Gao et al., 2021). Industrial Ubiquity quantifies the number of regions where a particular industry is already specialised, while Regional Diversity accounts for the diversity of industrial mixes within regions.

Details of the summary descriptive statistics, including Variance Inflation Factors (VIF), and pairwise correlation coefficients are provided in Supplemental Appendix B.1. Our analyses confirm that issues of multicollinearity and endogeneity do not distort our estimations.

Model specification

We adopt a fixed-effect regression approach by estimating entry and exit in a Linear Probability Model (LPM).⁸

Y_{i, p, t} = [E n t r y_{i, p, t}, E x i t_{i, p, t}]

(7)

\begin{array}{l} Y_{i, p, t} = β_{1} S R D_{i, p, t - 1} + β_{2} R D C o n t r o l s_{i, p, t - 1} \\ + β_{3} D e m o g r a p h i c C o n t r o l s_{i, p, t - 1} \\ + β_{4} I n d u s t r y C o n t r o l s_{i, p, t - 1} \\ + β_{5} G e o D i v e r s i f i c a t i o n_{i, p, t - 1} \\ + ρ_{p} + ι_{i} + γ_{t} + ε_{i, p, t} \end{array}

(8)

The principal coefficient of interest $β_{1}$ captures the impact of SRD on industries’ likelihood of either developing or losing a comparative advantage in manufacturing within a given region. We include three-way fixed effects for industries, regions, and time to account for time-independent characteristics that may correlate with the variables under study (Ganau, 2015; Gao et al., 2021). Additionally, we include control variables to address potential endogeneity and reduce the likelihood of reverse causality. We follow well-established practice in the diversification literature by lagging all independent variables by one period, denoted as t − 1 (Kogler et al., 2023; Neffke et al., 2011).

Results and discussion

We begin by estimating our models using the pooled dataset of 166 ITL-3 regions. The regression results are reported in Table 1, and the corresponding coefficient plots are presented in Figure 1. The first point to notice from Models 1–2 and 5–6 is a consistently significant positive (negative) correlation between higher SRD and the probabilities of industrial entry (exit) across both skill domains. This observation aligns with existing literature suggesting that regions are more likely to develop new specialisations in industries that require similar skill sets to those already embedded in their local workforce (Boschma et al., 2014; Corradini and Vanino, 2022; Neffke et al., 2017).

Table 1.

LPM regression results: Pooled models.

	Entry				Exit
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
TD SRD	0.0027*** (0.0003)		−0.0001 (0.0005)	0.0019*** (0.0004)	−0.0016*** (0.0002)		−0.0002 (0.0003)	−0.0012*** (0.0003)
SC SRD		0.0031*** (0.0003)	−0.0012* (0.0006)			−0.0017*** (0.0002)	0.0004 (0.0004)
TD × SC SRD			0.0002*** (0.0000)				−0.0001*** (0.0000)
LBR				0.0181 (0.0574)				−0.1559*** (0.0555)
TD SRD × LBR				0.0013** (0.0005)				−0.0006* (0.0004)
GERD	0.0410* (0.0248)	0.0391 (0.0248)	0.0402 (0.0247)	0.0413* (0.0248)	−0.0035 (0.0255)	−0.0026 (0.0255)	−0.0029 (0.0255)	−0.0036 (0.0255)
Pop. Density	0.0010 (0.0010)	0.0011 (0.0010)	0.0010 (0.0010)	0.0010 (0.0010)	−0.0008 (0.0010)	−0.0009 (0.0010)	−0.0008 (0.0010)	−0.0008 (0.0010)
Reg. Diversity	0.0086*** (0.0007)	0.0085*** (0.0007)	0.0085*** (0.0007)	0.0085*** (0.0007)	−0.0031*** (0.0006)	−0.0030*** (0.0006)	−0.0030*** (0.0006)	−0.0031*** (0.0006)
Ind. Ubiquity	−0.0005 (0.0004)	−0.0007 (0.0004)	−0.0006 (0.0004)	−0.0005 (0.0004)	0.0026*** (0.0004)	0.0027*** (0.0004)	0.0027*** (0.0004)	0.0026*** (0.0004)
Ind. Growth	−0.0088 (0.0089)	−0.0077 (0.0089)	−0.0076 (0.0089)	−0.0085 (0.0089)	0.0120 (0.0089)	0.0114 (0.0089)	0.0115 (0.0089)	0.0119 (0.0089)
Ind. Death	0.0048 (0.0106)	0.0042 (0.0106)	0.0029 (0.0106)	0.0045 (0.0106)	0.0006 (0.0104)	0.0009 (0.0104)	0.0015 (0.0104)	0.0007 (0.0104)
N	24,829	24,829	24,829	24,829	24,829	24,829	24,829	24,829
R ²	0.047	0.046	0.052	0.047	0.034	0.034	0.036	0.034
Adj. R²	0.039	0.038	0.044	0.039	0.026	0.026	0.028	0.026

Robust standard errors clustered at the region and industry level are in parentheses. All specifications include fixed effects for region, industry, and time.

GERD: Gross Domestic Expenditure on Research and Development; LPM: Linear Probability Model; SC: social-cognitive; TD: technical-digital.

p < 0.10. **p < 0.05. ***p < 0.01.

Figure 1.

Coefficient plots for the key variables.

To examine the interactive effects of TD SRD and SC SRD, we introduce the interaction term TD × SC SRD (Model 3 and 7, Table 1). The coefficients are statistically significant for both industrial entry (Model 3) and exit (Model 7). To aid interpretation in terms of probability changes, Figure 2(a) and (b) present the estimated marginal effects. A 1 SD increase in TD SRD is associated with a 22.45%⁹ increase in the probability of entry and a 14.26% reduction in the probability of exit. Similarly, a 1 SD increase in SC SRD corresponds to a 21.48% increase in entry and a 12.63% reduction in exit. Notably, the interaction term yields a 44.43% increase in the probability of entry and a 23.82% decrease in the probability of exit, substantially exceeding the effects of either main effect alone. This finding supports H1, indicating that regions with a balanced mix of related TD and SC skills are better positioned to achieve favourable diversification outcomes in manufacturing. The transformative potential of TD skills, through their role in adopting advanced Industry 4.0 technologies, is significantly amplified when combined with people-led SC skills, such as coordination, problem-solving, and adaptive expertise. This finding aligns with recent research emphasising cross-domain complementarities and synergies across different skill domains (Barzotto and De Propris, 2019; Corradini and Karoglou, 2022).

Figure 2.

Marginal effects. (a) Marginal entry effects (pooled models). (b) Marginal exit effects (pooled models). (c) Marginal entry effects (four regional groups). (d) Marginal exit effects (four regional groups).

To investigate whether TD SRD yields stronger diversification outcomes in left-behind regions, we introduce an interaction term between TD SRD and a binary indicator LBR¹⁰ (Model 4 and 8, Table 1). The interaction term TD SRD × LBR is positively associated with industrial entry and negatively with exit, indicating that TD SRD has a more pronounced effect on diversification in LBRs. Effect sizes in Figure 2(a) and (b) reveal that a 1 SD increase in TD SRD is associated with a 28.40% increase in the probability of entry in LBRs, compared to 14.07% in non-left-behind regions, approximately double the effect. Similarly, the reduction in the probability of exit is 17.12% in left-behind regions compared to 9.52% in non-left-behind regions. These results support H2 and align with previous research by Castellacci et al. (2019) and Barzotto et al. (2019), which suggests that left-behind regions possess greater scope to benefit from improvements in TD skills for forming new specialisations.

To further explore how relatedness within different skill domains and their interactions shape the place-based regional diversification patterns, we disaggregate the analysis by each of the four regional groups. The corresponding effect sizes are summarised in Figure 2(c) and (d), with full regression results reported in Supplemental Appendix B.2. The magnitudes of changes in probabilities across all groups consistently shows that TD × SC SRD has a significantly stronger effect on diversification than TD SRD’s main effect alone, aligning with the pooled regression results, further supporting H1.

Accounting for industrial heritage, we compare outcomes within regional groups that share similar historical manufacturing backgrounds. Among the two old industrial region types (LERs and LLRs), a 1 SD increase in TD SRD is associated with a larger increase in the probability of industrial entry in LERs (19.59%) compared to LLRs (18.91%), though it has a smaller effect on reducing exit. A similar pattern emerges in regions without a strong manufacturing legacy: LVRs exhibit a stronger entry effect than MDRs, but a more limited effect on exit. This divergence suggests that, within groups of regions with comparable industrial heritage, TD SRD has a stronger influence on industrial entry in left-behind regions, while its effect on reducing exit is comparatively weaker.

The findings on entry events are consistent with the pooled regression results, suggesting that the effect of TD SRD on industrial entry is more strongly shaped by regional development levels, irrespective of industrial heritage. In left-behind regions, where the existing skills base tends to be lower, improvements in TD SRD yield greater opportunities for industrial upgrading and new entry. The findings on exit events, however, provide further insights into how TD SRD shapes the place-based dynamics of regional diversification beyond what is captured in the pooled models, indicating that TD SRD’s influence on exit appears to be also conditioned by the impact of the regional industrial heritage. Specifically, in LERs, existing specialisations are often concentrated in traditional, low-value-added manufacturing industries with limited technological sophistication. Even as TD SRD improves, such industries may continue to decline due to their position in the industrial lifecycle, especially amid a broader structural shift toward advanced manufacturing (Sunley et al., 2021). This pattern may reflect the dynamics of industrial diversification as a process of creative destruction, whereby regional specialisations are gradually reconfigured: some incumbent industries decline, while capabilities are gradually developed that may support future entry into more technologically advanced activities (Deegan et al., 2024). In contrast, LVRs typically lack a significant manufacturing legacy; consequently, the very low baseline of existing specialisations may limit the scope for TD SRD to influence exit dynamics.

The impact of industrial heritage is also reflected in the role of SC SRD, albeit differently. In LERs, SC SRD’s impact is minimal both entry and exit outcomes, the stagnating industrial structures and entrenched networks create strong inertia, constraining combinatory and adaptive capacity (Roessler, 2025) and limiting the effectiveness of SC skills. This may also explain the interactive effect of TD × SC SRD: while the interaction term exhibits strong effects on industrial entry, its impact on reducing exits is far more limited, suggesting a constrained ability to retain existing specialisations even as new ones emerge in LERs. In contrast, in LVRs, SC SRD has a significantly stronger effect, enhancing both entry and exit outcomes, with the highest observed increase in entry probability at 25.14%. In these regions, characterised by a thin industrial base, diversification processes rely more heavily on SC skills, particularly those supporting the recombination of dispersed knowledge, the coordination of flexible production systems, and external engagement. However, the low level of regional development hinders the ability to leverage TD SRD, thereby constraining the potential of SC SRD to organise themselves into functioning social networks to turn into actual diversification. This echoes the challenges raised by Capello and Kroll (2016) and Iacobucci and Guzzini (2016) that these regions may find it difficult to engage in either scientific and technologically-based innovation (STI) or learning-by-doing, by-using, and by-interacting (DUI) mode of innovation. Taken together, the interplay of left-behind status and industrial heritage (whether overly entrenched or largely absent) constrains the potential for TD × SC SRD to generate meaningful diversification in both LERs and LVRs.

In comparison, we observe more robust diversification outcomes in LLRs and MDRs. In LLRs, both TD and SC SRD have the strongest observed impact on reducing industrial exits, and TD × SC SRD further amplifies both entry and exit. One explanation could be these regions have already undergone substantial restructuring of their industrial and organisational systems, potentially moving beyond the phase of creative destruction associated with traditional low-value-added manufacturing (a stage that LERs may still be navigating). Their existing industrial legacy may help lower search and coordination costs, facilitate strategic partnerships, and enable firms to identify and connect with complementary skills and technologies. In this context, skill relatedness may support not only the emergence of new regional specialisations but also the retention and reconsolidation of existing ones, which is vital for preserving regional resilience and enabling self-sustained growth (Fai et al., 2022).

Finally, we look at MDRs, where TD × SC SRD magnifies its complementary effect. One possible interpretation is that these regions may benefit from extensive STEM expertise combined with dynamic, highly skilled labour markets (Fitjar and Rodríguez-Pose, 2013). Such conditions may support more adaptive inter-organisational structures and network formations, potentially enabling greater experimentation with emerging technologies. In this context, despite lacking a dominant manufacturing heritage, the observed complementarity between TD and SC skill relatedness may reflect a synergy between advanced technological capabilities and more exploratory, experiential learning processes that support strong diversification dynamics. This pattern aligns with findings by Czaller and Hermann (2022) that large urban labour markets amplify the returns on SC skills, further strengthening diversification capacity in MDRs.

These findings across the four regional groups provide empirical support for H3, confirming that the effects of TD and SC skill relatedness (and their interaction) on regional diversification vary across regions with different development levels and industrial heritage.

Results from three types of robustness checks (see Supplemental Appendices C.1–C.3) are consistent with those reported in the main article, further reinforcing the robustness of our findings. A concise summary is provided in Figure 3.

Figure 3.

Summary of results.

Conclusion

This article investigates skill relatedness within two polarised skill domains across four UK regional groups characterised by distinct development levels and industrial heritages. We highlight three main results. First, there is a consistently significant positive association between higher levels of skill relatedness within both TD and SC domains and the likelihood of entry enhancement and exit reduction across all regional groups. Notably, the interaction between TD and SC skills exhibits a synergistic effect on industrial diversification, with SC SRD amplifying the impact of TD SRD on diversification. This points to the added value of integrating TD and SC skills in responding to the technological shifts transforming manufacturing (Acemoglu and Restrepo, 2019; Falck et al., 2021). Second, these effects are shown to be place-based. The impact of TD SRD on industrial entry is stronger in left-behind regions, regardless of whether they possess an industrial heritage, suggesting greater scope to benefit from improvements in TD skills in these regions (Barzotto et al., 2019; Castellacci et al., 2019). However, their relatively weaker influence on exit reduction highlights persistent challenges associated with different levels of regional industrial heritage. Third, regional development levels and the presence of industrial heritage jointly shape a region’s capacity to capitalise on the interaction between TD and SC SRD for effective industrial specialisations, leading to varied diversification patterns. Taken together, these findings advance the EEG agenda by offering novel insights into how relatedness within different skill domains and their interactions shape the place-based dynamics of regional diversification across different regional contexts.

Our findings carry important implications for regional policy. They provide a valuable perspective on how different regions diversify their manufacturing sectors by signalling demand for distinct combinations of related skills across two domains in the context of technological change, highlighting the need to embed skills policy more explicitly within targeted, place-based industrial strategies that leverage local strengths and address place-specific constraints (Bailey et al., 2023, 2026; Bathelt et al., 2024). Such strategies should be aligned with the existing local skills base and industrial heritage, and seek to cultivate a balanced combination of related TD and SC skills (Wang et al., 2026), thereby facilitating the emergence of new regional industrial specialisations. This is particularly relevant for left-behind regions, where targeted investments in related TD skills may yield higher marginal returns for industrial diversification and support processes of regional catch-up (Corradini et al., 2022; Giordano, 2020). At the same time, strengthening SC skills can enhance the effectiveness of TD skills, while also helping to mitigate disadvantages associated with persistent digital divides (Salemink et al., 2017).

We offer some caveats and avenues for future research. First, whilst our analysis focuses on relatedness, the literature suggests left-behind regions may also benefit from infusions of unrelated capabilities for more radical/less incremental breakthroughs (Antonietti and Montresor, 2021; Balland and Boschma, 2021; Boschma et al., 2023; Kogler et al., 2023). Investigating the effects of unrelated skills on regional diversification in future research may reveal alternative strategies to stimulate growth in left-behind regions where skill relatedness has not produced the strongest diversification outcomes (Breul et al., 2021; Frangenheim et al., 2019). Second, the broader constructs of skill complexity may also play crucial roles in shaping regional diversification (Buyukyazici et al., 2023). The absence of skill complexity considerations in the present analysis may result in an overestimation of the influence of relatedness alone. Future studies could integrate skill relatedness and complexity into a unified framework to further advanced EEG theorising. Third, due to the temporal scope of this study, we are unable to capture a truly long-term perspective in absolute terms. Future investigations could extend this research to encompass more extended timeframes, offering deeper insights into the sustained impacts of skill relatedness on regional diversification. Finally, although we examined the UK context, the frequent disjunction between skills policies and place-based industrial strategies is a common issue in many developed countries (OECD, 2017), so future research could explore the applicability of our findings beyond the UK.

Supplemental Material

sj-docx-1-eur-10.1177_09697764261451227 – Supplemental material for Skill relatedness: A dynamic approach to regional diversification in the UK

Supplemental material, sj-docx-1-eur-10.1177_09697764261451227 for Skill relatedness: A dynamic approach to regional diversification in the UK by Zihan Wang, Mariachiara Barzotto, Felicia M Fai and Philip R Tomlinson in European Urban and Regional Studies

Footnotes

Acknowledgements

We sincerely thank Prof. Ron Boschma, Prof Carlo Corradini, Prof. Elisa Giuliani, and Dr Dieter F. Kogler for their valuable comments and suggestions, as well as participants at the Bessemer Symposium in Economic Geography of Innovation (2024), II YSI – RSA Young Scholars’ Academy on Regional Studies workshop (2024), RSA Annual Conference (2024) for the feedback received on the early version of this paper.

ORCID iDs

Zihan Wang

Mariachiara Barzotto

Felicia M Fai

Philip R Tomlinson

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The work reported in this paper was undertaken as part of the Made Smarter Innovation: Centre for People-Led Digitalisation, at the University of Bath, University of Nottingham, and Loughborough University. The project is funded by the Engineering and Physical Sciences Research Council (EPSRC) Grant EP/V062042/1.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Supplemental material

Supplemental material for this article is available online.

Notes

References

Acemoglu

Autor

Hazell

Restrepo

(2022) Artificial intelligence and jobs: Evidence from online vacancies. Journal of Labor Economics 40(S1): S293–S340.

Acemoglu

Restrepo

(2018) Artificial intelligence, automation, and work. NBER Working Paper 24196.

Acemoglu

Restrepo

(2019) Automation and new tasks: How technology displaces and reinstates labor. Journal of Economic Perspectives 33(2): 3–30.

Alabdulkareem

Frank

Sun

AlShebli

Hidalgo

Rahwan

(2018) Unpacking the polarization of workplace skills. Science Advances 4(7): eaao6030.

Alan

Boneva

Ertac

(2019) Ever failed, try again, succeed better: Results from a randomized educational intervention on grit. The Quarterly Journal of Economics 134(3): 1121–1162.

Antonietti

Montresor

(2021) Going beyond relatedness: Regional diversification trajectories and key enabling technologies (KETs) in Italian regions. Economic Geography 97(2): 187–207.

Asheim

Grillitsch

Trippl

(2017) Introduction: Combinatorial knowledge bases, regional innovation, and development dynamics. Economic Geography 93(5): 429–435.

Autor

Dorn

Hanson

(2015) Untangling trade and technology: Evidence from local labour markets. The Economic Journal 125(584): 621–646.

Bacolod

Blum

Strange

(2009) Urban interactions: Soft skills versus specialization. Journal of Economic Geography 9(2): 227–262.

10.

Bailey

De Propris

Dimos

Fai

Hardy

Tomlinson

(2026) A critical review of the UK’s Modern Industrial Strategy: Lessons for ‘place-based’ policy. Regional Studies 60(1).

11.

Bailey

Pitelis

Tomlinson

(2023) Place-based industrial and regional strategy – Levelling the playing field. Regional Studies 57(6): 977–983.

12.

Balland

P-A

Boschma

(2021) Complementary interregional linkages and Smart Specialisation: An empirical study on European regions. Regional Studies 55(6): 1059–1070.

13.

Barzotto

Corradini

Fai

Labory

Tomlinson

(2019) Enhancing innovative capabilities in lagging regions: An extra-regional collaborative approach to RIS3. Cambridge Journal of Regions, Economy and Society 12(2): 213–232.

14.

Barzotto

Corradini

Fai

Labory

Tomlinson

(2020) Smart specialisation, Industry 4.0 and lagging regions: Some directions for policy. Regional Studies, Regional Science 7(1): 318–332.

15.

Barzotto

De Propris

(2019) Skill up: Smart work, occupational mix and regional productivity. Journal of Economic Geography 19(5): 1049–1075.

16.

Bathelt

Buchholz

Storper

(2024) The nature, causes, and consequences of inter-regional inequality. Journal of Economic Geography 24(3): 353–374.

17.

Bathelt

Storper

(2023) Related variety and regional development: A critique. Economic Geography 99(5): 441–470.

18.

Boschma

(2016) Relatedness as driver of regional diversification: A research agenda. Regional Studies 51(3): 351–364.

19.

Boschma

Eriksson

Lindgren

(2014) Labour market externalities and regional growth in Sweden: The importance of labour mobility between skill-related industries. Regional Studies 48(10): 1669–1690.

20.

Boschma

Miguelez

Moreno

Ocampo-Corrales

(2023) The role of relatedness and unrelatedness for the geography of technological breakthroughs in Europe. Economic Geography 99(2): 117–139.

21.

Breul

Hulke

Kalvelage

(2021) Path formation and reformation: studying the variegated consequences of path creation for regional development. Economic Geography 97(3): 213–234.

22.

Buyukyazici

(2024) Digital skills and industrial diversification of regions. Regional Studies, Regional Science 11(1): 583–598.

23.

Buyukyazici

Mazzoni

Riccaboni

Serti

(2023) Workplace skills as regional capabilities: Relatedness, complexity and industrial diversification of regions. Regional Studies 58(3): 469–489.

24.

Capello

Kroll

(2016) From theory to practice in smart specialization strategy: Emerging limits and possible future trajectories. European Planning Studies 24(8): 1393–1406.

25.

Castaldi

Drivas

(2023) Relatedness, cross-relatedness and regional innovation specializations: An analysis of technology, design, and market activities in Europe and the US. Economic Geography 99(3): 253–284.

26.

Castellacci

Consoli

Santoalha

(2019) The role of e-skills in technological diversification in European regions. Regional Studies 54(8): 1123–1135.

27.

Ciarli

Kenney

Massini

Piscitello

(2021) Digital technologies, innovation, and skills: Emerging trajectories and challenges. Research Policy 50(7): 104289.

28.

Cirillo

Evangelista

Guarascio

Sostero

(2021) Digitalization, routineness and employment: An exploration on Italian task-based data. Research Policy 50(7): 1–20.

29.

Clark

Bailey

(2018) Labour, work and regional resilience. Regional Studies 52(6): 741–744.

30.

Corradini

Karoglou

(2022) On the persistence and complementarities of design and technological change: A regional perspective. Regional Studies 57(7): 1289–1302.

31.

Corradini

Morris

Vanino

(2022) Towards a regional approach for skills policy. Regional Studies 57(6): 1043–1054.

32.

Corradini

Vanino

(2022) Path dependency, regional variety and the dynamics of new firm creation in rooted and pioneering industries. Journal of Economic Geography 22(3): 631–651.

33.

Czaller

Hermann

(2022) Return to skills and labour market size. Regional Studies 57(5): 800–813.

34.

De Propris

Bellandi

(2021) Regions beyond Industry 4.0. Regional Studies 55(10–11): 1609–1616.

35.

Deegan

Broekel

Haus-Reve

Fitjar

(2024) How regions diversify into new jobs: From related industries or related occupations? Regional Studies 58(11): 1965–1980.

36.

Deming

Kahn

(2018) Skill requirements across firms and labor markets: Evidence from job postings for professionals. Journal of Labor Economics 36(S1): S337–S369.

37.

Deming

Noray

(2020) Earnings dynamics, changing job skills, and STEM careers. The Quarterly Journal of Economics 135(4): 1965–2005.

38.

Eder

Trippl

(2019) Innovation in the periphery: Compensation and exploitation strategies. Growth and Change 50(4): 1511–1531.

39.

Eriksson

Hane-Weijman

(2016) How do regional economies respond to crises? The geography of job creation and destruction in Sweden (1990–2010). European Urban and Regional Studies 24(1): 87–103.

40.

Fai

Tomlinson

Branston

(2022) Actors, knowledge and path transformations in a declining cluster. European Urban and Regional Studies 29(4): 498–514.

41.

Falck

Heimisch-Roecker

Wiederhold

(2021) Returns to ICT skills. Research Policy 50(7): 1–17.

42.

Feldman

Kogler

(2010) Stylized Facts in the Geography of Innovation. Elsevier, pp.381–410.

43.

Fitjar

Rodríguez-Pose

(2013) Firm collaboration and modes of innovation in Norway. Research Policy 42(1): 128–138.

44.

Fitjar

Timmermans

(2016) Regional skill relatedness: Towards a new measure of regional related diversification. European Planning Studies 25(3): 516–538.

45.

Frangenheim

Trippl

Chlebna

(2019) Beyond the single path view: Interpath dynamics in regional contexts. Economic Geography 96(1): 31–51.

46.

Frenken

Van Oort

Verburg

(2007) Related variety, unrelated variety and regional economic growth. Regional Studies 41(5): 685–697.

47.

Frey

Osborne

(2017) The future of employment: How susceptible are jobs to computerisation? Technological Forecasting and Social Change 114: 254–280.

48.

Galetti

JRB

Tessarin

Morceiro

(2021) Skill relatedness, structural change and heterogeneous regions: Evidence from a developing country. Papers in Regional Science 100(6): 1355–1377.

49.

Ganau

(2015) Productivity, credit constraints and the role of short-run localization economies: Micro-evidence from Italy. Regional Studies 50(11): 1834–1848.

50.

Gao

Jun

Pentland

Zhou

Hidalgo

(2021) Spillovers across industries and regions in China’s regional economic diversification. Regional Studies 55(7): 1311–1326.

51.

Garcia-Lazaro

Mendez-Astudillo

Lattanzio

Larkin

Newnes

(2025) The digital skill premium: Evidence from job vacancy data. Economics Letters 250: 112294.

52.

Giordano

(2020) Post-Brexit regional economic development policy in the UK? Some enduring lessons from European Union Cohesion Policy. European Urban and Regional Studies 28(1): 26–33.

53.

Glaeser

Ponzetto

GAM

Tobio

(2012) Cities, skills and regional change. Regional Studies 48(1): 7–43.

54.

Grillitsch

Asheim

Trippl

(2018) Unrelated knowledge combinations: The unexplored potential for regional industrial path development. Cambridge Journal of Regions, Economy and Society 11(2): 257–274.

55.

Hassink

(2010) Locked in decline? On the role of regional lock-ins in old industrial areas. In: Boschma

Martin

(eds) The Handbook of Evolutionary Economic Geography, pp.450–470.

56.

Hassink

Gong

(2019) Six critical questions about smart specialization. European Planning Studies 27(10): 2049–2065.

57.

Henning

Eriksson

Garefelt

Martin

Elekes

(2025) Job relatedness, local skill coherence and economic performance: A job postings approach. Regional Studies, Regional Science 12(1): 95–122.

58.

Hidalgo

Klinger

Barabasi

Hausmann

(2007) The product space conditions the development of nations. Science 317(5837): 482–487.

59.

Iacobucci

Guzzini

(2016) Relatedness and connectivity in technological domains: Missing links in S3 design and implementation. European Planning Studies 24(8): 1511–1526.

60.

Isaksen

Tödtling

Trippl

(2018) Innovation policies for regional structural change: Combining actor-based and system-based strategies. In: Isaksen

Martin

Trippl

(eds) New Avenues for Regional Innovation Systems – Theoretical Advances, Empirical Cases and Policy Lessons. Springer, pp.221–238.

61.

Katz

Shapiro

(1985) Network externalities, competition, and compatibility. The American Economic Review 75(3): 424–440.

62.

Keeble

Walker

(2006) New firms, small firms and dead firms: Spatial patterns and determinants in the United Kingdom. Regional Studies 28(4): 411–427.

63.

Kogler

Whittle

Kim

Lengyel

(2023) Understanding regional branching: Knowledge diversification via inventor and firm collaboration networks. Economic Geography 99(5): 471–498.

64.

Le Petit-Guerin

Velthuis

Royer

Cauchi-Duval

Franklin

Leibert

Mackinnon

Pike

(2023) Lost in transition? Trajectories of regional “left-behindness” in the EU15 from 1982 to 2017. Beyond Left Behind Places Project Working Paper. Centre for Urban and Regional Development Studies (CURDS), Newcastle University.

65.

MacKinnon

Kempton

O’Brien

, et al. (2022) Reframing urban and regional “development” for “left behind” places. Cambridge Journal of Regions, Economy and Society 15(1): 39–56.

66.

Mandal

Gupta

(2016) A novel dictionary-based classification algorithm for opinion mining. In: 2016 Second International Conference on Research in Computational Intelligence and Communication Networks (ICRCICN). Kolkata, India, 23–25 September. IEEE, pp. 175–180.

67.

Martin

Sunley

(2006) Path dependence and regional economic evolution. Journal of Economic Geography 6(4): 395–437.

68.

Mayer

(2020) Slow innovation in Europe’s peripheral regions: Innovation beyond acceleration. Schlüsselakteure der Regionalentwicklung: Welche Perspektiven bietet Entrepreneurship für ländliche Räume 51: 9–22.

69.

McCann

(2019) Perceptions of regional inequality and the geography of discontent: Insights from the UK. Regional Studies 54(2): 256–267.

70.

Moretti

(2012) The New Geography of Jobs. Houghton Mifflin Harcourt.

71.

Morgan

Marques

(2019) The Public Animateur: Mission-led innovation and the “smart state” in Europe. Cambridge Journal of Regions, Economy and Society 12(2): 179–193.

72.

Morris

Vanino

Corradini

(2019) Effect of regional skill gaps and skill shortages on firm productivity. Environment and Planning A: Economy and Space 52(5): 933–952.

73.

Neffke

Henning

(2012) Skill relatedness and firm diversification. Strategic Management Journal 34(3): 297–316.

74.

Neffke

Henning

Boschma

(2011) How do regions diversify over time? Industry relatedness and the development of new growth paths in regions. Economic Geography 87(3): 237–265.

75.

Neffke

FMH

Otto

Weyh

(2017) Inter-industry labor flows. Journal of Economic Behavior & Organization 142: 275–292.

76.

OECD (2017) The Next Production Revolution: Implications for Governments and Business. OECD Publishing.

77.

OECD (2001) Understanding the Digital Divide. OECD Digital Economy Papers, No. 49. OECD Publishing. https://doi.org/10.1787/236405667766.

78.

Pinheiro

Hartmann

Boschma

, et al. (2022) The time and frequency of unrelated diversification. Research Policy 51(8): 104323.

79.

Plummer

Taylor

(2001) Theories of local economic growth (part 2): Model specification and empirical validation. Environment and Planning A: Economy and Space 33(3): 385–398.

80.

Poláková

Suleimanová

Madzík

, et al. (2023) Soft skills and their importance in the labour market under the conditions of Industry 5.0. Heliyon 9(8): e18670.

81.

Rayna

Striukova

(2021) Fostering skills for the 21st century: The role of Fab labs and makerspaces. Technological Forecasting and Social Change 164: 120391.

82.

Reveilhac

Morselli

(2022) Dictionary-based and machine learning classification approaches: A comparison for tonality and frame detection on Twitter data. Political Research Exchange 4(1): 1–25.

83.

Reveiu

Vasilescu

Banica

(2022) Digital divide across the European Union and labour market resilience. Regional Studies 57(12): 2391–2405.

84.

Rodríguez-Pose

(2018) The revenge of the places that don’t matter (and what to do about it). Cambridge Journal of Regions, Economy and Society 11(1): 189–209.

85.

Rodríguez-Pose

Crescenzi

(2008) Research and development, spillovers, innovation systems, and the genesis of regional growth in Europe. Regional Studies 42(1): 51–67.

86.

Rodríguez-Pose

Terrero-Dávila

Lee

(2023) Left-behind versus unequal places: Interpersonal inequality, economic decline and the rise of populism in the USA and Europe. Journal of Economic Geography 23(5): 951–977.

87.

Roessler

(2025) Unlocking green transitions: System-level agency in peripheral regions. Cambridge Journal of Regions, Economy and Society 19(1): 143–155.

88.

Salemink

Strijker

Bosworth

(2017) Rural development in the digital age: A systematic literature review on unequal ICT availability, adoption, and use in rural areas. Journal of Rural Studies 54: 360–371.

89.

Salvati

Carlucci

(2015) Patterns of sprawl: The socioeconomic and territorial profile of dispersed urban areas in Italy. Regional Studies 50(8): 1346–1359.

90.

Santoalha

Consoli

Castellacci

(2021) Digital skills, relatedness and green diversification: A study of European regions. Research Policy 50(9): 104340.

91.

Sunley

Evenhuis

Harris

, et al. (2021) Renewing industrial regions? Advanced manufacturing and industrial policy in Britain. Regional Studies 57(6): 1126–1140.

92.

Sunley

Martin

(2023) Place and industrial development: Paths to understanding? In: Bianchi

Labory

Tomlinson

(eds) Handbook of Industrial Development. Edward Elgar Publishing, pp.133–150.

93.

Tierney

Weller

Barnes

Beer

(2023) Left-behind neighbourhoods in old industrial regions. Regional Studies 58(6): 1192–1206.

94.

Tödtling

Trippl

(2013) Transformation of regional innovation systems: From old legacies to new development paths. In: Cooked

(ed.) Re-framing Regional Development. Routledge, pp.297–317.

95.

Tsvetkova

D’Amico

Lembcke

Knutsson

Vermeulen

(2024) “How well do online job postings match national sources in large English speaking countries?: Benchmarking Lightcast data against statistical sources across regions, sectors and occupations,” OECD Local Economic and Employment Development (LEED) Papers 2024/01. OECD Publishing.

96.

UKCES (2015) UKCES Employer Skills Survey 2015: UK report. UK Commission for Employment and Skills.

97.

Velthuis

Royer

Le Petit-Guerin

Cauchi-Duval

Franklin

Leibert

Mackinnon

Pike

(2024) Regional varieties of “left-behindness” in the EU15. Regional Studies 59(1): 1–16.

98.

von Hippel

(1994) “Sticky information” and the locus of problem solving: Implications for innovation. Management Science 40(4): 429–439.

99.

Wang

Fai

Tomlinson

Barzotto

Ehrhardt

(2026) Coworking spaces: A place-based approach for skill development in left-behind regions. Regional Studies 60(1): Article 2639508.

100.

Whittemore

(2018) Transversal competencies essential for future proofing the workforce. White Paper. Skillalibrary.

101.

Whittle

Kogler

(2020) Related to what? Reviewing the literature on technological relatedness: Where we are now and where can we go? Papers in Regional Science 99(1): 97–114.

102.

Xiao

Boschma

Andersson

(2018) Industrial diversification in Europe: The differentiated role of relatedness. Economic Geography 94(5): 514–549.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.17 MB

0.00 MB