The role of agglomeration in digitalisation and productivity: an empirical examination of manufacturing SMEs in South Korea

ABSTRACT

This study examines the relationship between regional agglomeration and the digitalisation of manufacturing small and medium enterprises (SMEs), and how this interplay affects their productivity. Using two analytical models, we investigate the correlation of regional agglomeration factors such as industrial specialisation, related and unrelated variety, urbanisation economies and the spatial concentration of suppliers with SMEs' digitalisation level in South Korea. Our findings indicate that related variety within regional agglomerations positively influences digital innovation, facilitating knowledge spillovers particularly beneficial for SMEs with higher levels of digitalisation. The study reveals a differentiated impact of agglomeration economies on SMEs based on their digital maturity. While high-level digitalised enterprises benefit from related variety and urbanisation economies, those in the initial stages of digital adoption gain proximity to digital technology suppliers. The 2SLS models further demonstrate that digitalisation significantly enhances productivity across all levels of digital technology adoption among SMEs. This study emphasises the conditional nature of the effects of regional agglomeration on digitalisation and productivity, advocating for nuanced support strategies that cater to the varying digital maturity stages of SMEs, which are valuable for policymakers and business leaders in fostering a supportive regional ecosystem that propels SMEs towards greater productivity in the digital economy.

KEYWORDS:

• Digitalisation
• agglomeration
• manufacturing
• innovation
• related variety

1. Introduction

The emergence of digital technologies and their application in manufacturing have attracted increasing attention as driving forces for transforming economic structures and bringing regional advantages. Digital technologies render production more flexible, autonomous, intelligent, and sustainable and reshape the business model (Frank et al., 2019; Mittal et al., 2018). This advancement is expected to restructure value chains and influence location decisions, including the reshoring of manufacturers in developed economies (Dachs et al., 2019; Rehnberg & Ponte, 2018). The recognition of the potential power of digital technologies has led to increasing research on the spatial perspective: geographical distribution of knowledge production in Industry 4.0 technologies (Balland & Boschma, 2021), the potential of digital urban production (Busch et al., 2021; Park, 2023), regional capability to integrate Industry 4.0 technologies (Muscio & Ciffolilli, 2020), spatial enterprises’ linkage to obtain Industry 4.0 technology and its performance (Barzotto & De Propris, 2021), the role of the technical knowledge-intensive business sector (Vaillant et al., 2021), spatial inequalities of digitalisation in small and medium enterprises (SMEs; Holl & Rama, 2023), and critical review of the reality of digital twin (Cooke, 2021). Although these pioneering studies examined the spatial aspects of how digital technologies are generated and used, few studies connected digitalisation to agglomeration, except for a conceptual framework on cluster and industry 4.0 (Götz & Jankowska, 2017) and an industry case study (Hervas-Oliver et al., 2019).

This study investigates the role of agglomeration in the context of manufacturing digitalisation and examines the effect of digitalisation on the productivity of SMEs. We consider the adoption of digital technology by manufacturing SMEs as an interactive and innovative activity. As SMEs typically do not have sufficient resources to develop their digital technology, they tend to rely on suppliers specialising in digital technology. Hence, not only does the application of digital technology in manufacturing SMEs require knowledge transfer between specialised suppliers and recipients, but it also requires engagement in on-site learning-by-doing practices that generate industry-specific knowledge on digitalising industrial data and using it to build more flexible and intelligent production systems. Therefore, we hypothesise that agglomeration promotes digital knowledge transfer within a region, increasing SME productivity. Specifically, this study investigates how regional agglomeration correlates with the digitalisation level of manufacturing SMEs by considering five measurements of agglomeration in each region such as industrial specialisation, related variety, unrelated variety, urbanisation economies, and the spatial concentration of suppliers in digital technology. It then examines how the digitalisation level correlates with the productivity of manufacturing SMEs. This empirical analysis contributes to the literature on innovation policy and regional science by theoretically integrating the concept of digitalisation into localised learning and knowledge transfer and empirically demonstrating the importance of space in the transformative era moving toward the digital economy. Our findings empirically demonstrate that agglomeration facilitates knowledge transfer in the digitalisation of manufacturing SME, albeit in varied ways contingent upon the stage of digitalisation maturity. In the early digitalisation stages, SMEs benefit from the spatial proximity to digital technology suppliers. However, as digitalisation matures, the significance of diverse and cognitively connected industrial structures and urbanised environments become pronounced in facilitating the adoption of digital technologies. Furthermore, our results consistently indicate that integrating digital technologies into manufacturing SMEs enhances productivity and imply that policymakers need to design digitalisation policies while considering variations in agglomeration.

2. Literature review

2.1. Manufacturing digitalisation

A substantial amount of literature has recognised digital technologies’ potential to change an enterprise’s strategy, organisational structure, production processes, skills, and business models (Büchi et al., 2020). The digitalisation of manufacturing is a data-driven transformation in processing activities, the way products are offered, and workers perform their jobs (Frank et al., 2019). Historically, advances in computer and information technologies such as computer-integrated manufacturing (CIM), computer-aided design (CAD), manufacturing execution systems (MES), and enterprise resource planning (ERP) have paved the way for automated production and vertical integration of organisations. However, information systems have remained isolated from other systems and have been under-analysed for effective applications. The recent rise in digital technologies such as the Internet of Things (IoT), cloud computing, and big data analytics has provided more cost-effective and flexible data collection, storage, processing, and analytical solutions that allow manufacturers to benefit from effective data use (Tao et al., 2018).

Data-driven transformations and the application of new digital solutions are expected to bring new characteristics to manufacturing operations and create new business models. These include (1) customer-centred production based on demand analysis and decentralised modular production systems (Friedrich et al., 2015); (2) a cyber-physical system operating in a physically networked virtual manufacturing environment consisting of industrial IoT and computational components (Gilchrist, 2016); (3) enhanced automation with humans using a collaborative robot, augmented and virtual reality device, and human-machine interface (Evjemo et al., 2020); and (4) creating a novel business opportunity by bundling product and service where a sensor, control system, and software are embedded, forming smart product-service systems (Chowdhury et al., 2018). These features enable manufacturers to improve their performance in various ways. The integration of digital technologies provides transparency to decision-makers, improves efficiency by optimising resources and energy, reduces the defect rate through real-time monitoring for product quality control, leads to more reliable processes with predictive equipment maintenance, and responds to demographic changes in the workplace by enhancing interactive collaboration between humans and machines (Kagermann et al., 2013). In summary, the efficient use of material, energy, and human resources in production; enhanced organisational capacity in terms of information sharing and product and production planning; and the introduction of new service products will improve the performance of manufacturers.

Along with envisioning the future of manufacturing, several studies empirically examined the state-of-the-art adoption and implementation of digital technologies (Frank et al., 2019; Mittal et al., 2018), the influence of internal and external factors (Agostini & Nosella, 2020; Raj et al., 2020; Won & Park, 2020), and perceived benefits (Büchi et al., 2020; Dalenogare et al., 2018). The literature suggests the following findings and implications. First, SMEs experience problems with the introduction of new digital technologies. SMEs have failed to identify concrete fields of action, programmes, and projects (Schumacher et al., 2016). According to an empirical analysis of France and India (Raj et al., 2020), the prominent problems faced by SMEs are a lack of digital strategies and resource scarcity. Many SMEs lack sufficient knowledge in the domain of digital technologies and cannot develop a vision or roadmap to guide their investment. This restricts the transition to digitalised products and production practices. Second, the adoption and diffusion of digital technology occur in stages according to the features of SMEs. An examination of Brazilian corporations showed that vertical integration – and monitoring-related digital technologies are widely employed by low-level adopters, that automation – and virtualisation-related technologies are introduced by moderate adopters, and digital technologies promoting flexible production are employed by the most advanced adopters (Frank et al., 2019). Third, organisational capability is a key factor in accounting for different adoption levels. Employee expertise, active involvement, manager commitment and support are key elements of organisational readiness (Won & Park, 2020). Internal and external cooperation is another aspect of an organisation’s capability. As SMEs face uncertainty and lack knowledge in the field of digital technologies, they need a strong partnership with suitable external service providers as well as intra-organisational cooperation among expert staff. Hence, it is important to develop soft skills and the ability to communicate, cooperate, and establish social connections (Agostini & Nosella, 2020). Fourth, emerging digital technologies are considered promising for improving SMEs’ performance. As the types of benefits differ according to the type of digital technology, SMEs cautiously decide to invest in the application of digital technologies for their purposes (Dalenogare et al., 2018). The existing literature documents the types of digital technologies introduced in SMEs in developing and developed countries, and examined the factors that play a role in facilitating SMEs’ adoption of digital technology. Although organisational absorptive capacity, which involves collaboration with external partners, is emphasised in the adoption of digital technologies, the spatial dimension and knowledge transfer have not been explicitly examined.

2.2. Role of agglomeration in manufacturing digitalisation and productivity

The literature posits that agglomeration is instrumental in facilitating knowledge spillover and innovation, particularly in the digitalisation of manufacturing. The dissemination of knowledge within these clusters occurs through a dynamic interplay between explicit and tacit forms, with the latter necessitating social interaction for transfer (Cowan et al., 2000; Vissers & Dankbaar, 2013). The typology of knowledge, as delineated by Lundvall and Johnson (1994), frames this exchange, in which the know-who and know-how, often resistant to codification, are accessed through interpersonal engagement. Cohen and Levinthal’s (1990) concept of an enterprise’s absorptive capacity emphasises that innovation hinges not on the ubiquity of knowledge, but on its accessibility within certain professional or geographical realms. Boschma (2005) critically appraised the role of proximity, suggesting that knowledge generation and transfer are more effectively promoted through cognitive, organisational, social, and institutional proximities than through mere geographical closeness. This expanded notion of proximity is supported by Storper and Venables (2004) and Malmberg and Maskell (2006), who advocate for the integral role of face-to-face interaction in the socialisation of the knowledge community and localised learning mechanisms that occur through the co-location of enterprises.

Digitalisation in manufacturing has emerged as a combinatorial innovation process that relies on the fusion of diverse knowledge bases from various engineering fields and the integration of general-purpose digital technologies into specific industrial applications (Strambach & Klement, 2012). The heart of this process is the interactive learning that unfolds between digital technology suppliers and manufacturing SMEs, where technology is not only applied but also adapted through practice and collaboration.

This synthesis of knowledge and practice is influenced by the spatial characteristics of agglomeration. The extant literature on spatial agglomeration suggests several forms that enhance innovation, particularly digitalisation, a critical component of modern innovation processes. Central to this discourse is the concept of industrial specialisation, which denotes the geographic clustering of enterprises within the same industry sector (Arrow, 1962; Glaeser et al., 1992; Marshall, 1961; Romer, 1986). Such specialisation can forge a concentrated nexus of sector-specific resources, expertise, and knowledge, which is crucial for manufacturing SMEs to access and deploy specialised digital tools and services effectively.

To further dissect the agglomeration phenomenon, studies have distinguished between related and unrelated variety as pivotal to knowledge recombination and innovation propagation (Aarstad et al., 2016; Antonietti & Cainelli, 2011; Frenken et al., 2007). Related variety represents a tapestry of cognitively linked industrial sectors within a region, enabling knowledge diffusion, which is particularly beneficial for manufacturing SMEs seeking to integrate digital innovations from adjacent sectors into their operations (Castaldi et al., 2015; Park & Choi, 2022). Conversely, unrelated variety, characterised by a broad spectrum of non-cognate industries, introduces a wealth of diverse ideas and technologies that potentially drive radical innovation and digital solution development (Castaldi et al., 2015; Park & Choi, 2022).

Urbanisation economies constitute another dimension, encapsulating the advantages enterprises gain from their presence in large and densely populated urban areas endowed with advanced infrastructure such as high-speed internet and digital services, which are indispensable tools for digitalisation (Batty et al., 2012; Neirotti et al., 2014). The resultant proximity to a diverse talent pool and cross-pollination with educational and research institutions in such settings provides manufacturing SMEs with access to cutting-edge digital R&D, thereby fostering a culture of digital innovation (Glaeser, 2011).

Finally, the spatial concentration of suppliers, particularly those specialising in digital technologies, emerges as a unique form of agglomeration. This proximity fosters horizontal links between suppliers and manufacturing SMEs, optimising the supply chain, and reducing transaction costs. The co-location of these entities facilitates a more efficient exchange of goods, services, and knowledge, which is essential for the digital transformation process (Baptista & Swann, 1998). By minimising the costs associated with coordination and communication, the spatial concentration of suppliers significantly influences the rate and success of digital adoption among SMEs, suggesting that this proximity is particularly beneficial during the early stages of digital integration (Boschma, 2005; McCann & Simonen, 2005).

Agglomeration’s impact extends to productivity. Enterprises’ proximity leads to efficient resource sharing, collective learning, and innovation, all of which are productivity catalysts. Regions with high agglomeration often outperform productivity measures, reflecting synergistic benefits (Melo et al., 2009; Rosenthal & Strange, 2004). Digitalisation magnifies these productivity gains. Manufacturing processes have become more efficient and flexible by integrating digital technologies, yielding significant improvements in productivity (Brynjolfsson & Hitt, 2000; Haller & Siedschlag, 2011).

In summary, as discussed above, a large body of literature observed regionally bounded innovative activities and accounted for why a region matters in knowledge spillover. They showed that regional industrial structure, measured by concentration or relatedness, would be influential in facilitating innovative activities. However, the role of agglomeration for regionally bounded innovation has not yet been addressed in the digital transformation context. Our study contributes to regional innovation literature in two main points. First, we theoretically connect manufacturing digitalisation, agglomeration, and spatially bounded knowledge spillover. Manufacturing digitalisation is a combinational form of knowledge between the digital field and domain fields of manufacturing. The lack of resources for manufacturing SMEs requires partnerships with suppliers specialized in a digital field. The adopter’s condition, such as cognitive relatedness and proximity among entities, affects the spread of digitalisation. We posit that a certain regional industrial structure may create a conducive environment to spread digital technologies in a region. Second, we provide an empirical framework to measure the digitalisation of manufacturing SMEs and analyse what types of regional industrial structure matter in digital technology spread and adoption of manufacturing SMEs.

3. Case selection, data, and methods

3.1. Case study area

We chose South Korea as a case study area because the country has a strong manufacturing base and has implemented a proactive policy for manufacturing digitalisation called the Manufacturing Innovation 3.0 Strategy (Wiktorsson et al., 2018). This initiative stems from the recognition that while South Korean SMEs are competitive in their respective industrial domains, they often face challenges owing to a lack of specialised knowledge in digital technologies. As a remedy, the government has championed digitalisation strategies across different industrial sectors, emphasising partnerships with specialised suppliers. A unique aspect of South Korea’s industrial landscape is the intentional placement of core industries because of the government’s location policy, which has led to marked regional disparities in industrial structures and concentrations. Recent empirical studies suggest that regional variations in industrial structures and concentrations directly affect an enterprise’s innovation capabilities and overall productivity (Choi & Choi, 2017; Park & Choi, 2022). Thus, it would be interesting to explore the effects of these different characteristics of agglomeration on the digitalisation of manufacturing SMEs across regions in South Korea.

3.2. Data

The data used in this study were raw data from the 2018 Survey on the Information Level of Korean Small and Medium Enterprises offered by the Ministry of SMEs and Startups of Korea and Korea Technology and Information Promotion Agency for SMEs. The survey began in 2002 to understand IT usage in the industry, and recently its scope was extended to the areas of digital technology and smart factories. As the data cover all industry sectors, we selected SMEs in the manufacturing sector and excluded cases that did not contain sales and profit information. We used 1,707 manufacturing SMEs in South Korea.

3.3. Methods

To explore the relationships between ‘agglomeration and digitalisation level’ and ‘digitalisation and productivity’, we employed two analytical models: multi-level regression and two-stage least squares (2SLS) models. Initially, our investigation concentrated on assessing the level at which regional agglomeration exerts a favourable impact on the digitalisation level of manufacturing SMEs. For this analysis, variables influencing an enterprise’s digitalisation level were systematically categorised and scrutinised at both enterprise and regional levels. Thus, the multi-level regression model is suitable for handling this nested or hierarchically structured dataset (e.g. enterprise, regional, and national level; Hox et al., 2017). Subsequently, 2SLS techniques were implemented to further probe the contribution of digitalisation levels in manufacturing SMEs to productivity enhancement. Given the potential simultaneous endogeneity between the digitalisation level and productivity, the 2SLS technique was adopted to provide consistent estimators, functioning as a method of instrumental variable estimation.

The variables for the analytical models were measured as follows. First, to compute the digitalisation level of SMEs as a key dependent variable, we considered two aspects: the scope of work areas to which digital technology is applied and number of new digital technologies being adopted. Two questions in the survey were important. The first question was whether an information system is established in the work area, which consisted of 13 detailed operational activities in sales, procurement, production, logistics, and administration. The second question concerns whether emerging digital technologies are being introduced in an enterprise. It includes nine digital technologies: cloud computing, big data analytics, IoT, artificial intelligence, blockchain, online to offline, 3D printing, virtual and augmented reality, and software-driven solutions. As there are 22 individual question items related to digital technology adoption, the digitalisation level score in SMEs ranges from 0 to 22. If information technologies are widely used and a large number of emerging digital technologies are introduced, a higher digitalisation score is assigned to that enterprise. The level of digitalisation was normalised to the z-score when used as a dependent variable. Second, regarding productivity as another key dependent variable, we measured total factor productivity (TFP). Enterprise-level productivity was estimated using a Cobb–Douglas production function, with employees and capital in the survey dataset as a factor endowment.

Third, regional agglomeration, as a key explanatory variable, was calculated at the municipal level (called si-gun-gu). As detailed in the previous section, agglomeration variables, which influence innovation and digital technology adoption in manufacturing SMEs, were measured along five dimensions that are supported by robust economic theories and empirical evidence. The first is the industrial specialisation of the region, which is a proxy for spatially bounded industry-specific knowledge spillover. The locational quotient (LQ) of the manufacturing sector, based on the two-digit Korean Standard Industry Classification, was used to calculate the industrial specialisation of the region.¹ The second and third are related and unrelated variety, which test whether knowledge of manufacturing digitalisation is more likely to diffuse in the cluster of enterprises with cognitively close or diverse non-cognate industry sectors, respectively. Related and unrelated variety were assessed using the method described by Frenken et al. (2007).² The fourth category is urbanisation economies, which were calculated using the population size of each municipality. We assume that urbanisation economies in populous urban areas provide SMEs with advanced infrastructure, diverse talent, educational resources, fostering digitalisation, and competitive technological innovation. The last is the geographic concentration of digital technology suppliers. As we assume that cooperation between digital technology suppliers and manufacturing is a channel of knowledge transfer, the co-location of specialised suppliers and manufacturing SMEs is a relevant factor. The number of digitalisation technology suppliers in each municipality was used to identify the effect of geographical proximity between suppliers and recipients.

Finally, we consider control variables at the enterprise level, such as digitalisation level, productivity, size, internal R&D organisation, exports, and service products, which are potentially associated with the digitalisation levels of manufacturing SMEs. We expected productivity, size (measured by sales), and exports to be positively associated with the digitalisation level due to the abundance of internal resources. As internal R&D organisations may indicate an enterprise’s innovativeness, they are more likely to accept emerging technologies. We also considered that manufacturing servitisation would be relevant to digitalisation. If a manufacturing enterprise provides service products, it may show a higher level of digitalisation.

Additionally, we consider the absorptive capability of SMEs as a control variable at the enterprise level, as recent studies have suggested that absorptive capability is a key component that accounts for the level of digitalisation in manufacturing SMEs (Agostini & Nosella, 2020; Raj et al., 2020; Won & Park, 2020). Absorptive capability pertains to individual employee expertise and collective organisational efforts such as on-site job training and information sharing. We built three explanatory variables representing the SMEs’ absorptive capabilities retrieved from the survey questionnaire: the number of employees specialising in information and digital technology, the number of employees participating in job training related to information and digital technology, and the organisational innovation effort in promoting digitalisation. The first two variables were measured by counting the number of employees, and the third was based on a question asked to evaluate the organisational innovation effort in promoting digitalisation on a five-point Likert scale.

The descriptive statistics of the variables are presented in Table 1. Some explanatory variables such as sales, urbanisation, and digital technology suppliers were converted to natural logarithms and input into the models. Stata 16/MP 16.1 was used for multi-level regression and 2SLS techniques.

Table 1. Descriptive statistics.

Variable	Obs.	Mean	Std. dev.	Min.	Max.
Dependent variable
Digitalisation level	1707	0.000	1.000	−1.660	1.940
Productivity	1707	1.837	0.179	1.190	2.442
Independent variable
Enterprise level (Level 1)
Sales	1707	9.401	1.498	5.011	14.301
R&D organisation (dummy)	1707	0.528	0.499	0.000	1.000
Export (dummy)	1707	0.414	0.493	0.000	1.000
Service product (dummy)	1707	0.006	0.080	0.000	1.000
Employees specialising in digital technology	1707	2.071	2.814	0.000	30.000
Employees participating in job training	1707	26.643	104.566	0.000	2000.000
Effort of organisational innovation	1707	2.481	0.973	1.000	5.000
Municipal level (Level 2)
Specialisation	1707	3.235	5.116	0.023	73.555
Related variety	1707	0.860	0.199	0.000	1.411
Unrelated variety	1707	2.503	0.237	1.328	3.138
Urbanisation economies	1707	12.649	0.788	9.981	13.993
Suppliers in digital technology	1707	1.303	2.610	−4.605	4.913

Note: Number of municipalities = 175.

Figure 1 illustrates the spatial patterns of key variables across regions in South Korea. The symbology for each map has been classified using natural breaks (Jenks) method. Variables of digitalisation level and productivity, measured at the enterprise level, are represented by mean values for each municipality (si-gun-gu). These variables exhibit higher values in regions with a strong manufacturing base, notably in the northwestern and southeastern parts of the country. Among the variables measured at the municipal level, related variety displays a geographic pattern similar to the aforementioned variables, whereas industrial specialisation does not exhibit a notable regional disparity. Additionally, suppliers in digital technology and urbanisation economies have higher values in metro cities. Such spatial patterns imply a potential correlation between the digitalisation and productivity of SMEs and regional agglomeration factors, notably related variety. The statistical relationship between these variables will be explored in the following section.

Figure 1. Spatial patterns of digitalisation level, productivity, and regional agglomeration.

4. Results and discussion

4.1. Agglomeration and digitalisation level of manufacturing SMEs

Table 2 presents the results of the multi-level regression models analysing the correlation between regional agglomeration and the digitalisation level of manufacturing SMEs in South Korea, with standard errors in parentheses adjacent to the coefficient estimates. Model 1 in Column 1 of Table 2 presents the outcomes for all manufacturing SMEs. At the municipal level, Model 1 indicates that only related variety exhibits a statistically significant positive coefficient, suggesting that regional agglomeration economies, specifically related variety, play a facilitative role in the digitalisation of manufacturing SMEs. This form of agglomeration is conducive to innovation through cross-industry collaboration and knowledge spillover. Manufacturing SMEs in such areas can benefit from digital innovations originating in adjacent sectors, which can be adapted or integrated into their operations.

Table 2. Factors associated with digitalisation level of manufacturing SMEs.

	Model 1 (All manufacturing SMEs)		Model 2 (Low digitalisation level manufacturing SMEs)		Model 3 (High digitalisation level manufacturing SMEs)
Enterprise level (Level 1)
Productivity	0.969***	(0.213)	0.310*	(0.183)	0.875***	(0.148)
Sales	0.122***	(0.024)	0.064***	(0.021)	0.005	(0.016)
R&D organisation (dummy)	0.119***	(0.045)	0.115***	(0.041)	0.016	(0.030)
Export (dummy)	0.085*	(0.045)	0.043	(0.042)	0.032	(0.028)
Service product (dummy)	−0.242	(0.244)	−0.185	(0.187)	−0.059	(0.208)
Employees specialising in digital technology	0.093***	(0.008)	0.041***	(0.015)	0.028***	(0.004)
Employees participating in job training	−0.000	(0.000)	−0.000	(0.000)	−0.000	(0.000)
Effort of organisational innovation	0.114***	(0.021)	0.017	(0.020)	0.073***	(0.013)
Municipal level (Level 2)
Specialisation	0.005	(0.004)	0.004	(0.004)	0.003	(0.002)
Related variety	0.246**	(0.121)	0.136	(0.109)	0.191**	(0.079)
Unrelated variety	−0.160	(0.098)	−0.049	(0.090)	−0.052	(0.063)
Urbanisation economies	0.032	(0.040)	−0.026	(0.036)	0.071***	(0.026)
Suppliers in digital technology	−0.004	(0.012)	0.018*	(0.010)	−0.025***	(0.008)
Cons.	−3.730***	(0.572)	−1.789***	(0.502)	−2.046***	(0.395)
Number of observations	1707		889		818
Groups	175		175		175
ICC	0.0021		0.0078		0.0000
AIC	4132.114		1407.789		671.7673
BIC	4219.194		1484.431		747.0771

Note: standard errors in parentheses.

* p < 0.1, ** p < 0.05, *** p < 0.01.

Models 2 and 3, detailed in Columns 2 and 3 of Table 2, segregate manufacturing SMEs based on digitalisation levels, distinguishing between those in the lower 50% (Model 2) and upper 50% (Model 3). The results of these models reveal that the type of regional agglomeration exerts a variegated impact on the digitalisation levels of manufacturing SMEs. A positive and statistically significant correlation with related variety was observed solely among high digitalisation level enterprises in Model 3, whereas this significance was absent for low digitalisation level enterprises in Model 2. Additionally, Model 3 indicates that high-level enterprises benefit from the positive effects of urbanisation economies on their digitalisation endeavours. In contrast, in Model 3, a negative association is reported between high-level enterprises and the spatial concentration of digital technology suppliers, whereas in Model 2, this spatial concentration is positively impactful for low-level enterprises. These insights indicate that regional agglomeration economies influence the digitalisation of manufacturing SMEs in a differentiated manner contingent upon their digital maturity. Specifically, the cognitively interconnected diversity of regional industrial structures (related variety) and sophisticated infrastructure and innovation networks in urban conurbations (urbanisation economies) are conducive to the digitalisation of enterprises with a high level of digital adoption. However, these agglomeration economies are less impactful for enterprises with lower levels of digitalisation. Conversely, the regional concentration of digital technology suppliers exerts a positive influence exclusively on enterprises with lower levels of digitalisation, underscoring the pivotal role of these suppliers in the nascent stages of digital adoption.

In Model 1, concerning enterprise-level variables, several of these factors are significant and positively correlated with the digitalisation level of manufacturing SMEs, with the exception of employee participation in job training and the presence of a service product. Specifically, the variables indicative of an SME’s productivity, sales performance, internal R&D departments, and engagement in international markets demonstrate a positive relationship with the adoption of advanced digital technology. Notably, these results suggest that manufacturing SMEs with higher productivity, extensive sales, in-house R&D capabilities, and presence in foreign markets are more inclined to integrate sophisticated digital technologies. However, contrary to expectations, the variable representing the presence of a service product is negative and lacks statistical significance. This implies that the current landscape of digital technologies is predominantly implemented within the domains of production and logistics, rather than in the service sector.³ With respect to the absorptive capacity of SMEs, two variables are statistically significant: the number of employees specialising in digital technology and degree of organisational effort toward innovation are positively correlated with SME’ digitalisation levels. In contrast, employee participation in job training does not significantly impact digitalisation. This may suggest that while specific investments in digital skills and innovation are conducive to digital adoption, general employee training and service-oriented product offerings do not necessarily predict or facilitate the digitalisation level within SMEs.

4.2. Digitalisation and productivity of manufacturing SMEs

Table 3 presents the results of the 2SLS models that examine how the level of digitalisation in manufacturing SMEs correlates with their productivity. Row 1 of Table 3 systematically identifies the factors that influence productivity. In Model 1, the digitalisation level emerges as a variable with a statistically significant and positive coefficient, suggesting that the integration of digital technologies by manufacturing SMEs correlates positively with productivity enhancements. This trend remains consistent across Models 2 and 3, indicating that, despite variations in the extent of digital technology adoption, the incorporation of digital tools and processes universally promotes productivity improvements within manufacturing SMEs.

Table 3. Factors associated with digitalisation level and productivity of manufacturing SMEs.

	Model 1 (All manufacturing SMEs)		Model 2 (Low digitalisation level manufacturing SMEs)		Model 3 (High digitalisation level manufacturing SMEs)
Equation for productivity
Digitalisation level	0.022***	(0.003)	0.014**	(0.006)	0.065***	(0.008)
Sales	0.086***	(0.002)	0.088***	(0.003)	0.083***	(0.003)
R&D organisation (dummy)	0.045***	(0.005)	0.052***	(0.007)	0.037***	(0.007)
Export (dummy)	0.009*	(0.005)	0.003	(0.008)	0.015**	(0.007)
Constant	0.999***	(0.017)	0.979***	(0.024)	0.988***	(0.024)
Equation for digitalisation level
Sales	0.202***	(0.016)	0.091***	(0.014)	0.074***	(0.012)
R&D organisation (dummy)	0.167***	(0.044)	0.133***	(0.039)	0.053*	(0.029)
Export (dummy)	0.094**	(0.045)	0.045	(0.042)	0.048*	(0.029)
Service product (dummy)	−0.303	(0.245)	−0.206	(0.187)	−0.050	(0.213)
Employees specialising in digital technology	0.101***	(0.008)	0.043***	(0.015)	0.034***	(0.004)
Employees participating in job training	0.000	(0.000)	−0.000	(0.000)	0.000	(0.000)
Effort of organisational innovation	0.117***	(0.021)	0.018	(0.020)	0.075***	(0.013)
Specialisation	0.006	(0.004)	0.004	(0.004)	0.003	(0.002)
Related variety	0.237**	(0.119)	0.148	(0.105)	0.150*	(0.080)
Unrelated variety	−0.188*	(0.097)	−0.068	(0.086)	−0.055	(0.064)
Urbanisation economies	0.030	(0.039)	−0.024	(0.034)	0.064**	(0.027)
Suppliers in digital technology	−0.004	(0.012)	0.017*	(0.010)	−0.024***	(0.008)
Constant	−2.646***	(0.516)	−1.478***	(0.444)	−0.994***	(0.363)
Number of observations	1707		889		818
R-squared
Equation for productivity	0.7276		0.6708		0.7078
Equation for digitalisation level	0.3450		0.1204		0.2986

Note: standard errors in parentheses.

* p < 0.1, ** p < 0.05, *** p < 0.01.

Regarding the analysis of the factors correlated with the level of digital technology adoption, as depicted in Row 2 of Table 3, a notable parallel to the insights gleaned from the multi-level analysis in Table 2 exists. Model 1 of Table 3 posits a robust positive correlation between related variety and digitalisation, whereas it documents a negative relationship with unrelated variety. This distinction emphasises the critical importance of related variety, which refers to the cognitively interconnected diversity within regional industrial structures, in stimulating innovation and advancing the digitalisation agenda in manufacturing SMEs.

Expanding on this, Models 2 and 3 offer a nuanced segmentation of SMEs based on their digitalisation status. These models show that enterprises boasting a higher digitalisation stature benefit substantially from the related variety and dynamics of urbanisation economies. However, these elements have a diminished influence on enterprises in the lower spectrum of digitalisation. Additionally, strategic positioning relative to digital technology suppliers manifests differential effects, providing a competitive edge to enterprises with a budding level of digitalisation, while this benefit is not as pronounced for enterprises that have already achieved a higher tier of digital integration.

5. Conclusion

This study investigates the intricate web of factors that drives digitalisation and productivity among manufacturing SMEs in South Korea, highlighting the influential role of regional agglomeration. Our study revealed that the proliferation of digital technologies is not an isolated process but is significantly shaped by the regional characteristics of agglomeration. The main findings are as follows:

First, central to our findings is the pivotal role of related variety in regional agglomerations, which is particularly salient for enterprises with advanced digitalisation. This interconnection of industry diversity creates a fertile environment for digital innovation, as evidenced by the thriving high digitalisation-level enterprises in such clusters. These enterprises benefit from the cross-pollination of ideas and shared innovation endemic to these regions. Conversely, the lack of a substantial correlation between related variety and digitalisation in less digitally mature enterprises implies that other, more dominant obstacles to digital adoption that are not alleviated by industry diversity alone exist.

Next, our study shows the differential impacts of urbanisation economies and proximity to digital technology suppliers, suggesting a progression in the digital maturation of manufacturing SMEs. Early stage enterprises seem to derive significant benefits from accessibility to digital technology suppliers, who likely provide critical support and technological access. By contrast, this proximity appears to confer no additional advantages to highly digitalised enterprises, suggesting that they may have reached a level of digital maturity where direct access to technology suppliers is no longer a constraint.

Additionally, our analysis consistently demonstrates that digitalisation acts as a catalyst for productivity, a trend that persists regardless of the level of digital adoption. This finding corroborates the assertion that digital technology integration serves as a cornerstone in enhancing the efficiency and performance of SME manufacturing.

In summary, the empirical evidence gathered advances our understanding of the dynamic interplay between regional agglomeration patterns and enterprise-level characteristics, and how they collectively influence the digitalisation trajectory of manufacturing SMEs. The conditional nature of these influences is underscored, with the impact of various factors waxing or waning depending on the digital maturity of enterprises. This study provides valuable insights into the tailored strategies needed to support SMEs at different stages of their digitalisation journey, ultimately fostering productivity growth and innovation.

This study’s findings have several policy implications. As governments recognise the potential for innovation stemming from the combination of digital technology in manufacturing, they persevere with initiatives to promote digital convergence in manufacturing, including the Advanced Manufacturing Partnership in the U.S., Made in China 2025 in China, Future of Manufacturing in the U.K., Impresa 4.0 in Italy, and Manufacturing Innovation 3.0 Strategy in South Korea (Büchi et al., 2020). A place-based policy approach would be beneficial, owing to the features of localised learning and knowledge spillover in digitalisation (Götz & Jankowska, 2017). Rather than providing a subsidy to an individual enterprise, a policy initiative should focus on identifying industry-specific needs and providing support to develop an applicable solution and practical know-how, and then disseminate information on innovative models and know-how by boosting industrial networking opportunities. Policymakers must design industrial digitalisation policies to leverage the agglomeration economy.

This study left the following research topics unexplored. Although this study confirms the impact of the agglomeration economy on digitalisation, it does not explicitly examine the channels through which knowledge spreads and localised learning occurs. Perhaps the anchor institutions leading innovative applications are key actors, and imitation or adaptation occurs through formal and informal networks around the anchor institution. Investigating key actors and their local and global networks to identify channels for knowledge diffusion is a fruitful approach. Second, our analysis reveals that the agglomeration of related variety would be more influential in more mature SMEs of digitalisation. Considering the different impacts and diverse needs of digital technologies (Frank et al., 2019), in-depth industrial case study is necessary. A case study by Hervas-Oliver et al. (2019) is an example exploring the collective efforts of local actors transitioning toward digitalisation. The question of what sorts of SME’s capacities are crucial and how these capabilities are associated with agglomeration should be explored through in-depth case studies. Finally, although this study focused on measuring the total factor productivity as an impact of manufacturing SMEs’ digitalisation, digitalisation affects corporate performance in more diverse ways. While digitalisation has short-term impacts, such as efficient resource use and defect reduction, it has a more profound impact on manufacturing as a change in the business model based on product servitisation. This analysis attempted to consider service products; however, meaningful results were not obtained, likely due to the lack of a mature business model. Examining the impact of advanced digitalisation and associated changes in business models on the regional economy would be desirable.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by National Research Foundation of Korea [grant number: NRF-2022R1A2C2004671].

Notes on contributors

Taelim Choi

Taelim Choi is a research fellow at the Incheon Institute. His research focuses on sustainable manufacturing, regional economic and environmental modelling, local labour markets, the digital economy and regional economic development policy.

Jeong-Il Park

Jeong-Il Park is an associate professor at Keimyung University in South Korea. He studies sustainable local and regional economic development planning, with a focus on digital transformation, urban manufacturing, urban industrial spaces and green industries.

Notes

1 The LQ is computed as $\mathrm{LQ}={\displaystyle \frac{\mathrm{Fir}{\mathrm{m}}_{\mathrm{gr}}/\mathrm{Fir}{\mathrm{m}}_r}{\mathrm{Fir}{\mathrm{m}}_g/\mathrm{Fir}{\mathrm{m}}_{\mathrm{nation}}}}$. g refers to the industry sector, and r refers to the region.

2 Related variety is $\mathrm{Related}\mathrm{variety}=\sum _{g=1}^G{p}_g{H}_g$, $H_g=\sum _{i=1}^I{p}_{\mathrm{ig}}\ln \left({\displaystyle \frac{1}{p_{\mathrm{ig}}}}\right)$. Unrelated variety is $\mathrm{Unrelated}\mathrm{variety}=\sum _{g=1}^G{p}_g\ln \left({\displaystyle \frac{1}{p_g}}\right).$ g refers to the category of the parent industry, i is a subgroup category within the parent industry, and r refers to the region.

3 The average digitalisation scores in the four subcategories are 1.06 for sales, 1.10 for procurement, 2.43 for production and logistics, and 1.84 for administration. The digitalisation level of production and logistics is roughly twice that of the other three subcategories.