The interactions between digitalization, innovation and employment in European companies: Insights from a Latent Class Analysis

Vodă, Adina-Maria; Ciobotea, Mihai; Badea, Doina; Roman, Monica; Stan, Marian Article The interactions between digitalization, innovation and employment in European companies: Insights from a Latent Class Analysis Economies Provided in Cooperation with: MDPI – Multidisciplinary Digital Publishing Institute, Basel Suggested Citation: Vodă, Adina-Maria; Ciobotea, Mihai; Badea, Doina; Roman, Monica; Stan, Marian (2025) : The interactions between digitalization, innovation and employment in European companies: Insights from a Latent Class Analysis, Economies, ISSN 2227-7099, MDPI, Basel, Vol. 13, Iss. 4, pp. 1-19, https://doi.org/10.3390/economies13040104 This Version is available at: https://hdl.handle.net/10419/329384 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Academic Editor: Tsutomu Harada Received: 28 February 2025 Revised: 31 March 2025 Accepted: 2 April 2025 Published: 8 April 2025 Citation: Vodă, A.-M., Ciobotea, M., Badea, D., Roman, M., & Stan, M. (2025). The Interactions Between Digitalization, Innovation and Employment in European Companies: Insights from a Latent Class Analysis. Economies,13(4), 104. https://doi.org/ 10.3390/economies13040104 Copyright: © 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/ licenses/by/4.0/). Article The Interactions Between Digitalization, Innovation and Employment in European Companies: Insights from a Latent Class Analysis Adina-Maria Vodă, Mihai Ciobotea , Doina Badea * , Monica Roman and Marian Stan Department of Statistics and Econometrics, Bucharest University of Economic Studies, 010371 Bucharest, Romania; [email protected] (A.-M.V.); [email protected] (M.C.); [email protected]o (M.R.); marian.stan.r[email protected] (M.S.) *Correspondence: [email protected] Abstract: There is increasing concern regarding the association between technological change and jobs. This study explores how different patterns of digitalization and innovation relate to job creation in European companies. We use data from the European Company Survey 2019 collected by Eurofound and Cedefop. We apply Latent Class Analysis (LCA) to identify the typologies of companies, mainly based on their level of technology adoption, innovation practices and employment patterns. We showcase four distinct classes of companies: moderate adoption of digital technology and strong international orientation, traditional and local, medium digitalization, process innovative with local focus and digital leaders and innovators, with specific patterns regarding digitalization, innovation and job creation. The digital leaders and innovators class revealed a high level of digitalization and innovation and maintained stable employment levels, with increased investments in staff training and tendency towards automation. Conversely, less-digitalized traditional companies are more susceptible to stagnation or employment decline. In general, the employment outlook is stable, without significant employment growth, signaling the need for balanced investments in innovation and digitalization that stimulate more and better jobs. This is the first study to apply LCA to explore complex relationships between digitalization, innovation, foreign trade, training investments and employment trends and offers fresh insights into company views towards employment in the digital era. Keywords: digitalization; employment; human capital; innovation; latent class analysis; job creation 1. Introduction Over the last 20 years, the rapid pace of technological change has transformed labor markets, reshaping skill requirements and workforce dynamics in an unprecedent way. Nonetheless, the impact of technology adoption on workforce demand at the company level can result in job loss due to automation. However, it can also create new jobs through the uptake of new technologies, development of new products and services and increased demand for products and services, or due to the emergence of heavy industries (Acemoglu & Restrepo,2020). As new technologies become much more affordable, organizations remained competitive, yet their ability to align workforce capabilities with these technological demands varies significantly from country to country, from one sector of activity to another, and from company to company. Although existing research has addressed the effects of digitalization Economies 2025,13, 104 https://doi.org/10.3390/economies13040104 Economies 2025,13, 104 2 of 19 on labor markets, there is a gap in understanding the association between company-level technology adoption and employment outcomes. This study aims to address this gap in research by using latent class analysis (LCA) to identify the distinct typologies of companies based on their level of technology adoption, innovation practices, employment patterns and other key characteristics. LCA is a powerful methodological instrument for revealing hidden subgroups within a population, thus allowing us to explore simple correlations to identify complex, multidimensional patterns of behavior within European companies. Drawing on data from the European Company Survey (2019), conducted by the European Foundation for the Improvement of Living and Working Conditions (Eurofound) and European Center for the Development of Vocational Training (Cedefop), our analysis is based on 21,869 managers’ responses to a standardized survey from subsidiary sites of multi-site companies across the EU-27 and the United Kingdom. This rich dataset provides valuable understanding of employment patterns, innovation strategies and approaches to skill development and digitalization, offering a great outlook into European companies’ behaviors in the digital era. Our research questions are the following: Q1. What are the key typologies of companies that can be identified based on their approach to digitalization, innovation and international market exposure and combined characteristics such as size, sector and country of residence? Q2. How do the identified typologies interact with the companies’ approaches to present and future employment changes and staff training? Since the beginning of the First Industrial Revolution in 1750, jobs and people’s lifestyles have been threatened by machines, which have the potential to replace them. Job loss and the lack of the capacity to adapt to technological advancements are among the greatest fears individuals face (Kozak et al.,2020). Every industrial revolution has brought high uncertainty and fears related to jobs becoming obsolete. John Maynard Keynes even predicted that jobs automation would lead to the famous ‘technical unemployment’ (Keynes,1931). Nevertheless, these fears have been rapidly abandoned, as evidence shows that technology and innovation have great potential to create new industries and jobs while making older sectors obsolete. The scientific literature was dominated by these concepts, built around the well-known theories of ‘creative destruction’ (Schumpeter,1942) and ‘compensation theory’ (Marx,1867). Within the context of the Fourth Industrial Revolution, which brought about an exponential pace of technological development, various questions arose: Is this time any different? Will robots steal jobs (Acemoglu & Restrepo,2020)? Will advances in artificial intelligence and automation lead to important job losses (e.g., Brynjolfsson & McAfee,2014; Ford,2015)? The emerging scientific literature highlights that organizations with higher levels of technology adoption often demonstrate a greater propensity for increased productivity, innovation and workforce skilling (Bartel et al.,2007). Conversely, low levels of technological integration have been associated with job polarization, favoring the allocation of low-paid routine tasks to the service sector (Autor & Dorn,2013;Obadic,2020). While the future of labor is uncertain (World Bank,2019), research highlights that the increased use of digital technologies in companies is associated with a growth in the number of employees or positive future employment prospects (García-Romanos & MartínezRos,2024). Various scientific opinions have emerged in this regard. On the one hand, according to creative destruction theory, innovation is considered to eliminate old jobs but creates new, higher-value roles, resulting in a net positive impact (Mastrostefano & Pianta, 2009). Likewise, innovation leads to job creation through increased firm productivity and Economies 2025,13, 104 3 of 19 expansion (Blanchflower & Burgess,1999;Van Reenen,1997). In addition, the rise of digital industries creates jobs in new sectors; thus, digitalization has contributed to the emergence of entirely new industries, creating millions of jobs in fields such as programming and data analysis (Acemoglu & Restrepo,2020). Finally, digitalization is flexible, facilitating job creation across diverse industries, including services, by enabling efficiency and new tasks (Brynjolfsson & McAfee,2014). For instance, digitalization has a small but positive net effect on employment when it is evaluated across the German economy using structural models (Arntz et al.,2016). More specifically, non-machine-based digital technologies such as Enterprise Resource Planning (ERP) and e-commerce positively affect employment by enabling task efficiency and creating complementary human roles (Aubert-Tarby et al., 2018). However, digitalization has the potential to influence the workforce structure differently at the company level. That is, increased investment in digitalization is associated with increased employment of highly skilled workers and reduces employment of lowskilled workers (Autor & Dorn,2013;Balsmeier & Woerter,2019). Thus, automation and digitalization are unlikely to result in substantial job losses. Workers with lower qualifications may need to adapt more because of their higher risk of automation. The main challenge remains regarding the capacity to address rising inequality and ensure worker retraining (Arntz et al.,2016). Further research reveals that innovation at the firm level, including digitalization, generally leads to employment growth by creating new job opportunities (Blanchflower & Burgess,1999). Innovation, particularly through research and development (R&D) investments, fosters job creation within firms and across industries, contributing to positive employment effects (García-Romanos & Martínez-Ros,2024;Van Reenen,1997). This belief continues to dominate the concept of employment growth over time. Ten years later, based on Schumpeter’s theory of “creative destruction”, it was argued that innovation, including digitalization, continues to have a net positive impact on jobs at the industry level (Mastrostefano & Pianta,2009). However, firm strategies continue to be essential, as product innovations associated with digitalization often have more positive employment effects than process innovations (Mastrostefano & Pianta,2009). The positive synergies between product and process innovations, digital technology adoption and use and the learning capacity of the organization have been largely explored by researchers over the last half of the century. Adopting innovation boosts both product and process improvements, enhancing firm competitiveness through digital technologies that streamline production and improve customer engagement (Rogers,1983;Tornatzky & Klein,1982). Digital technology can undoubtedly promote important processes and product innovation at the company level (Ayoko,2021). Digital technologies also facilitate organizational learning, which is important for the implementation of new processes. Technologies such as ERP systems and electronic data interchange assist organizations in developing dynamic capabilities, allowing them to create new workflows and products (Cohen & Levinthal,1989,1990). Bibliometric analysis concludes that digital innovation adoption is prevalent in industries focusing on product and process advancements, while high R&D industries, such as technology and manufacturing, heavily adopt digital tools, leading to innovative products and process redesign (Van Oorschot et al.,2018). More recently, analyses conducted at the EU level have shown that investments in digital technology and learning capacities can positively influence product and process innovation and innovation overall (Curzi & Ferrarini,2024;Greenan & Napolitano,2024). Nevertheless, it is widely accepted that increased productivity at the company level is achieved through investments in digital technology and employee training (Chen et al., 2016;Siegenthaler & Stucki,2015). Economies 2025,13, 104 4 of 19 Digitalization creates a landscape in which larger firms and tech-savvy industries lead the charge. Larger companies and those in technology-intensive sectors (e.g., information and communications) are more likely to adopt digitalization (e.g., e-commerce, robots, or data analytics). Undoubtedly, company size matters for larger firms, and exporters demonstrate higher levels of digitalization, suggesting a correlation between firm size and technology adoption. Larger firms are more likely to adopt digital technologies. For instance, in Switzerland, adoption rates are notably higher among firms with more than 20 employees (Balsmeier & Woerter,2019). The size of an organization, as well as its readiness, affects the adoption of digital technology and its impact on innovation. The presence of barriers to innovation and digitalization in small and medium enterprises negatively affects efficiency (Mitropoulos et al.,2024). However, larger organizations and those with higher technological readiness are more inclined to adopt digital tools that contribute to innovative processes and products (Damanpour & Schneider,2006). Large firms in advanced industries prefer machine-based digital technologies that combine data access, computation and hardware. Smaller firms and less advanced sectors focus more on non-machine-based technologies (Aubert-Tarby et al.,2018;Arntz et al.,2016). Moreover, industry type is equally important: manufacturing sectors, which are more technologically advanced, adopt complex digital technologies more frequently than the service sectors. For example, complex digital technologies are associated with sectors that require high technical sophistication. Technologies such as robotics are primarily adopted in sectors with significant manufacturing or technological components, highlighting the role of industry type in digitalization levels (Balsmeier & Woerter,2019). Nevertheless, technological advancements have an important influence on knowledge-intensive and highly technologized sectors, and conversely, a negative influence on employment growth in less technologically intensive manufacturing sectors (Obadic,2020). Additionally, competitiveness stimulates digitalization, as exposure to international competition stimulates firms in export-driven industries to embrace digital technologies. The level of digitalization within a firm’s industry is closely linked to its competitiveness. Switzerland’s significant exposure to international competition stimulates firms, especially those in the manufacturing and export-dominant sectors, to embrace digital technologies more extensively. This indicates that an industry’s competitiveness plays a crucial role in determining its degree of digitalization (Bris & Cabolis,2017). When focusing on interregional trade in Europe, researchers found that routine-replacing technological change had an important displacement effect on jobs during 1999 and 2010 and also led to net growth in job creation due to increased product demand (Gregory et al.,2022). Human capital plays a significant role, as high salaries and skills in advanced industries support the adoption of digital technology. Switzerland’s skilled workforce and high salaries have led firms in advanced industries to invest in digital technologies to enhance efficiency and reduce costs. This trend is particularly noticeable in larger companies and those in the manufacturing sector (Siegenthaler & Stucki,2015). From an employment perspective, the skill level of an employee is particularly important, as IT can be a substitute for, but can also complement, human labor in Europe (Peng & Zhang,2020). The economic theories presented above suggest that the relationship between digitalization, innovation and employment outcomes operates through a layered mechanism. Technological inputs—such as the adoption of robots, e-commerce platforms and specialized software and IT systems, along with product and process innovation—are first correlated as firms’ internal capacities and strategies. These inputs are linked to how companies organize production, introduce new products and services to the market and adapt their workforce. In our research, we use LCA to capture this multidimensional complexity Economies 2025,13, 104 5 of 19 by identifying latent company, based on observable patterns in firms’ technology adoption and innovation behavior. These profiles—ranging from digitally advanced companies to traditional, low-innovation firms—are used as an analytical lens through which we explore how European organizations respond strategically to digital transformation. Although the LCA method does not allow us to establish causal relationships, we use it as proxy to understand the economic adjustment mechanisms underlying company-level responses such as recent employment changes, future employment expectations and training investment. We use this as an organizational mediator to link innovation and digitalization to employment outcomes such as stability, job growth or decline. This approach integrates the theoretical perspectives on the creative destruction economic model into an empirical framework that reflects the variations in firm behavior. Fueled by renewed European Union (EU) interest to rebuild the EU companies’ competitiveness globally by funding investments that should close the innovation gap with China and United States of America, especially by relaunching advanced technologies and closing the skills gap, the present analysis could contribute to better understanding of where EU innovative and digitalized companies stand in terms of employment prospects and what can be done further (Draghi,2024). 2. Materials and Methods The data source for responding to the research questions was the European Company Survey (ECS) for the year 2019, conducted by Eurofound in collaboration with Cedefop (Eurofound & Cedefop,2020). ECS 2019 focuses on workplace organization, innovation and companies’ approaches to skills and digitalization. Data were collected from 21,869 subsidiary sites of multi-sites companies across Europe, covering all 27 EU member states and the UK. The target population of the survey was companies with more than ten employees. The data refer to managers’ responses to the survey. ECS 2019 employs a push-to-web methodology combined with fieldwork. Companies were initially contacted by phone to designate a senior manager for the survey. The designated individuals were invited to complete an online questionnaire. The purpose of the ECS 2019 edition was to analyze workplace organization and innovation at the company level, as well as companies’ approaches to skills and digitalization (Eurofound & Cedefop,2020). A multistage stratified random sampling method was used, and procedures varied across countries. The sample was stratified by company size (number of employees) and sector of activity (manufacturing, construction and services). This approach aimed to ensure high-quality data and to collect representative data at both the national and EU levels. A total of 21,869 interviews were conducted with management personnel. To respond to the research questions, this study uses a comprehensive latent class analysis aimed at identifying latent employment behaviors in EU companies grouped mainly based on their level of technology adoption and innovation patterns. To explore how different patterns of digital technology adoption at the company level are associated with employment trends in the European Union (EU), we applied a Latent Class Analysis (LCA), considering its broad use in scientific research in multiple domains pertaining to social sciences research (Santos et al.,2015;Zhu,2024), psychology (Petersen et al.,2019) and health (Zhou et al.,2018;Kongsted & Nielsen,2017). The essence of the LCA method involves finding underlying groups in a population using categorical data based on probabilistic modeling, resulting in intra-classes homogeneity and inter-class heterogeneity. Starting from the two research questions (company typologies related to their approach to digitalization, innovation and international market exposure and companies’ approach to present and future changes in employment and staff training), this research aimed to identify groups in firm populations with similar features. Economies 2025,13, 104 6 of 19 The dataset is peculiar, as the data are categorical. This disqualifies other methods designed for use on numerical datasets, such as cluster analysis (deterministic method used for continuous numerical data) or neural network analysis for clustering (which is adequate for analyzing large datasets with many numerical variables and complex nonlinear relationships). LCA was utilized, since it is a modern, powerful modeling technique that responds to the typology of the variables (qualitative data resulting from a large survey) and the scope of the analysis. More specifically, to identify hidden company profiles, LCA proved helpful in revealing latent subgroups of companies that share similar patterns in technology adoption, innovation activities and employment behaviors, considering the type of ECS 2019 variables (mainly categorical). Traditional methods (e.g., linear regression) often mask such heterogeneity, whereas LCA helps uncover nuanced profiles (Weller et al.,2020). Similarly, LCA brings a multidimensional perspective to the analysis. As the dataset includes various measures (technology adoption, training, innovation and workforce evolution), LCA allows us to perform a simultaneous analysis of these variables. It groups companies with statistically similar response patterns into classes and provides a comprehensive view of how technology spreads across different company types. From a practical viewpoint, LCA results reveal new perspectives on employment trends in the EU that are useful to various stakeholders, such as policy makers or company management. By understanding the specific classes of companies, decision makers have better tools to adjust policies and management strategies to the specific needs of each profile, especially in the context of the overall behaviors uncovered by the LCA. Finally, we used LCA for innovation purposes, considering the uniqueness of the approach. While many studies on the intersection between digitalization and employment use descriptive statistics or econometric models, LCA is relatively underutilized to explore how technology shapes workforce dynamics at the company level across Europe. The method’s ability to segment companies based on multiple dimensions simultaneously provides novel insights and complements existing macroeconomic and case-based research. The variables used for the scope of this analysis (see Appendix A) were selected based on the outcomes of the literature review. Employment trends are measured using two ordinal variables that focus on how the total number of employees in the company evolved since the beginning of 2016 (CHEMP) and the evolution of the total number of employees in the company over the next three years (CHEMPFUT) (Acemoglu & Restrepo,2020). Professional training (organized inside or outside the company) offered by the company to employees during paid working hours was captured by an ordinal variable (PAIDTRAIN). Digitalization in companies was measured using specific ECS 2019 ordinal variables, showcasing the degree of digital adoption in the company. These variables are e-commerce (investigating whether the company buys or sells goods or services online (e.g., through business-to-business portals, e-commerce, etc.); robot usage (ICTROB, exploring whether the company uses robots, which in this context are defined as programmable machines that can perform a series of complex actions automatically, including interactions with humans); and usage of software designed or customized for its needs over the last three years (ICTAPP) (Acemoglu & Restrepo,2020;Aubert-Tarby et al.,2018;Brynjolfsson & McAfee,2014). Innovation at the company level was measured using two dichotomous variables, in line with other studies that have used the European Company Survey 2019 (Della Torre et al.,2021;Curzi & Ferrarini,2024), by focusing both on innovation in products (INNOPROD) and processes (INNOPROC) over a period of three years before the survey (starting in 2016) (Mastrostefano & Pianta,2009). Economies 2025,13, 104 7 of 19 Exposure to international markets was captured by an ordinal variable (SALESINT), which measures the percentage of sales to international clients over a three-year period starting in 2016 (Bris & Cabolis,2017). A set of variables regarding the company profile and country of residence was also used (see Appendix A), such as the size of the company (WPSIZE_MM_N), the main sector of activity following the NACE 2-Digit codes level (MAINACT) and country of residence, the latest covering the EU 27 Member States and United Kingdom (Aubert-Tarby et al., 2018;Arntz et al.,2016;Balsmeier & Woerter,2019). The Latent Class Analysis (LCA) was performed using the R programming language with a code designed by the authors. Subsequent steps were followed, starting with the preparation of the environment, including by loading the necessary libraries. Furthermore, the data were cleaned, and missing values were handled (by filling gaps in categorical variables using the mode and numerical variables using the mean). As LCA can only be performed on categorical data, the numerical variables were discretized. The variables were coded as factors to make them fit the statistical model. For the LCA modeling, we identified four underlying latent groups for the data considered. The LCA method is probabilistic, and it models the probability that an observation is in a certain latent class Ckon the basis of the formula below: P(X=x|Ck)= J ∏ j=1 PXj=xj Ck where P(X=x|Ck) represents the probability that a set of X answers will be included in a latent class C k ; in our case, it represents the probability that a company will be part of a C k class with the answers X = x ·∏J j=1 represents the product of the probabilities for each observed variable; we assume that the variables are independent inside a latent class. Xj=xj Ck represents the probability that variable X j will take the value x j , considering class C k . This shows how each variable contributes to a latent class definition. Additionally, we set up the data frames used in the model, as well as the three covariates pertaining to employment and training, and we encoded them as factors. Furthermore, we conducted an LCA analysis. Within the analysis, the Bayesian Information Criterion (BIC) for several scenarios considering two, three, four, five, six, or seven classes was also calculated. The number of classes was determined based on the BIC. The BIC can be calculated as follows: BIC =−2·ln(L)+k·ln(N) where: •ln(L): model log-likelihood; •k: number of parameters; •N: sample size. The BIC was chosen as the indicator because it strikes a balance between model fit and complexity (Wagenmakers,2007). The model with the lowest BIC was selected as the most suitable for interpretation (see Table 1). Economies 2025,13, 104 8 of 19 Table 1. BIC values. No. of Classes BIC Value 2 classes 443,551.8 3 classes 439,827.7 4 classes 438,033.4 5 classes 437,955.7 6 classes 452,663.7 7 classes 452,876.0 Source: Author’s own calculations. The quality of the clustering was ensured by selecting the number of classes based on the lowest BIC. The BIC was run for several possible numbers of clusters, resulting in the values given below (see Table 1): Although the BIC showed that the 5-class solution would be optimum from the method point of view, the team opted for a 4-class model for clearer interpretability and practical relevance for this research. The slight statistical improvement of the 5-class over the 4-class did not justify adding more complexity. Based on a comparison of the resulting BIC values, it was determined that the most appropriate number of classes was four. The results were then visualized through several graphical representations, such as the probabilities for each class, the heat map, the countries’ heat map and the covariate effect on class membership. In order to estimate the quality of the class clarity, the team calculated the entropy for this LCA model. The entropy was evaluated using the following formula (Celeux & Soromenho,1996): Entropy =1+∑N i=1∑K k=1Pik·lnPik +e−12 N·ln(K) where: •N: total number of observations. •K: total number of latent classes. The value of the entropy is between 0 and 1, with values closer to 1 indicating better classification. The calculated entropy was 0.6837. This value shows moderate clarity for the latent class model. There are several reasons for why this value is not higher, and perhaps the most important is that the BIC value for a 5-class model is very close. The calculated entropy if we consider 5 classes is 0.7275; close to the first 4-class situation (chosen by the team for analysis reasons). 3. Results The results of the LCA, considering the optimal BIC for choosing the number of classes, show that ECS 2019 companies are classified into four classes based on their latent profiles of digitalization and innovation determinants. When investigating the class membership distribution of the participants in the survey (See Figure 1), we observed that Class 3 was the most numerous one (7791 companies), followed by Class 2 (5038 companies), Class 4 (4934 companies) and Class 1 (3609 companies). Economies 2025,13, 104 15 of 19 shows a complex interaction linked to the level of technology and innovation adoption and key company characteristics. Class 1 and Class 3 companies demonstrated moderate levels of digitalization and the tendency to maintain stable employment trends. Conversely, Class 2 companies, characterized by low digitalization, displayed declining employment trends. The highly digitalized Class 4 companies presented signs of automation that could hamper employment growth in the future. Notably, none of these categories demonstrated a significant increase in job creation, thereby questioning the applicability of the primary theory of ‘creative destruction’, which has been highlighted by various authors throughout all technological revolutions. This study empirically traces the link between firms’ technological profiles, strategic behavior and employment outcomes. While LCA does not allow for causal inference, the consistent patterns observed across latent classes provide valuable insight into how firms adapt or fail to adapt to digital transformation. Our results indicate that digitalization and innovation, even when accompanied by training investment, do not automatically translate into job creation. The expected pattern of compensation, where job losses due to automation are compensated by new roles in digitally mature firms, is not clearly present. This has important implications for policy: if technological transformation remains confined to a narrow segment of firms and does not translate into broader employment gains, more targeted interventions may be required to ensure inclusive labor market outcomes in the digital era. In conclusion, the findings of the four latent class analyses highlight that public policies need to promote the benefits of digitalization while mitigating potential negative employment outcomes. Public policies should aim to support the efforts of Class 2 less digitalized, traditional companies, and to adopt digitalization and innovation strategies observed in more advanced firms, such as those in Classes 1 and 4, by offering financial incentives to encourage digital adoption and innovation (Crisan et al.,2023). Likewise, proactive measures to create diverse employment opportunities for jobseekers should be incentivized. Nevertheless, professional training should become a priority both for companies and governments, as informed by the success of Class 4 firms. Therefore, governmental funds, including European investments, should be directed towards upskilling, reskilling and improving the digital skills of European employees to increase their adaptability to labor markets (Caravella et al.,2023). While advanced digitalization does not necessarily reduce jobs, the observed behavior of Class 4 companies, which showed a tendency towards automation and training over workforce expansion, suggest the need for strategies that balance digitalization and automation with employment sustainability. This study has important limitations. These include the lack of longitudinal data and various macroeconomic variables. The data concerning future employment tendencies are based on HR representatives’ subjective responses and expectations, introducing a certain degree of subjectivity, while the potential impact of the COVID-19 pandemic was also missed. Nevertheless, other limitations stem from missing data on investments in R&D across Europe, productivity, technological investments and skill mismatch. Using LCA to identify specific features and typologies of digitalized companies in conjunction with employment is an innovative approach that should be further explored and replicated. Further research is essential to develop a deeper understanding of the underlying conditions under which digitalization leads to inclusive and sustainable economic growth and sustainable job creation. Conducting a similar analysis on the latest ECS data could help shape EU policies and investments in digitalization, innovation, training and employment programs, particularly in light of the Draghi report (Draghi,2024). Economies 2025,13, 104 16 of 19 Author Contributions: Conceptualization, A.-M.V. and M.C.; methodology, A.-M.V., M.C. and M.R.; software, M.C.; validation, M.R., A.-M.V. and M.C.; formal analysis, A.-M.V. and M.C.; investigation, A.-M.V., M.S. and D.B.; resources, A.-M.V., M.S. and D.B.; data curation, A.-M.V. and M.C.; writing— original draft preparation, A.-M.V.; writing—review and editing, A.-M.V., M.R., M.S., M.C. and D.B.; visualization, M.C. and A.-M.V.; supervision, A.-M.V.; project administration, A.-M.V.; funding acquisition, all authors. All authors have read and agreed to the published version of the manuscript. Funding: This research received no external funding. Institutional Review Board Statement: Not applicable. Informed Consent Statement: Not applicable. Data Availability Statement: The data used in this study can be accessed at UK Data Service: http://doi.org/10.5255/UKDA-SN-8691-1 (accessed on 24 May 2023). Conflicts of Interest: The authors declare no conflict of interest. Abbreviations The following abbreviations are used in this manuscript: ECS European Company Survey (2019) EU European Union UK United Kingdom LCA Latent Class Analysis R&D Research and Development Appendix A Table A1. ECS 2019 survey questions. Authors’ own selection. Description Item Scale and Details Variable Name Country of the establishment Country EU 27, United Kingdom countrycode Establishment’s main activity category Nace-1 digit codes The following sectors were excluded: - [1]—Agriculture, forestry and fishing - [15]—Public administration and defence; compulsory social security - [16]—Education - [17]—Human health and social work activities - [20]—Activities of households as employers; - [21]—Activities of extraterritorial organisations and bodies mainact_d How many people work in this establishment? 1–3 - Small (10–49 employees) - Medium (50–249 employees) - Large (250+ employees) mm_size_grp Economies 2025,13, 104 17 of 19 Table A1. Cont. Description Item Scale and Details Variable Name International sales Since the beginning of 2016, what percentage of this establishment’s sales were to customers in other countries? 1–4 - We do not engage in export (0%) - 1% to 24% - 25% to 49% - 50% or more salesint_cat Digitalization Does this establishment buy or sell goods or services on the internet? For instance, by using business-to-business portals, e-commerce etc. Yes/No ecommerce Since the beginning of 2016, did this establishment purchase any software that was specifically developed or customized to meet the needs of the establishment? Yes/No ictapp_cat Robots are programmable machines that are capable of carrying out a complex series of actions automatically, which may include the interaction with people. Does this establishment use robots? Yes/No ictrob_cat Innovation Since the beginning of 2016, has this establishment introduced any new or significantly changed processes either for producing goods or supplying services? 1–3 - Yes, new to the market - Yes, new to the establishment, but not new to the market - No innoproc Since the beginning of 2016, has this establishment introduced any new or significantly changed products or services? 1–3 - Yes, new to the market - Yes, new to the establishment, but not new to the market - No innoprod Training investments In 2018, how many employees in this establishment participated in training sessions on the establishment premises or at other locations during paid working time? (%) 1–7 - None at all - Less than 20% - 20% to 39% - 40% to 59% - 60% to 79% - 80% to 99% - All paid_train_d.1 Employment trends How has the total number of employees in this establishment changed since the beginning of 2016? 1–5 - Increased by more than 10% - Increased by up to 10% - Stayed about the same - Decreased by up to 10% - Decreased by more than 10% chemp_change_employees In the next three years, how do you expect the total number of employees in this establishment to change? 1–3 - It will increase - It will stay about the same - It will decrease chenmpfut_change_empl_3_years Economies 2025,13, 104 18 of 19 References Acemoglu, D., & Restrepo, P. (2020). Robots and jobs: Evidence from US labor markets. Journal of Political Economy,128(6), 2188–2244. Arntz, M., Gregory, T., & Zierahn, U. (2016). The risk of automation for jobs in OECD countries: A comparative analysis. OECD Social, Employment and Migration Working Papers, No. 189. Organisation for Economic Co-Operation and Development. Aubert-Tarby, C., Escobar, O. R., & Rayna, T. (2018). The impact of technological change on employment: The case of press digitization. Technological Forecasting and Social Change,128, 36–45. [CrossRef] Autor, D. H., & Dorn, D. (2013). The growth of low-skill service jobs and the polarization of the US labor market. American Economic Review,103(5), 1553–1597. [CrossRef] Ayoko, O. B. (2021). Digital transformation, robotics, artificial intelligence, and innovation. Journal of Management and Organization, 27(5), 831–835. [CrossRef] Balsmeier, B., & Woerter, M. (2019). Is this time different? How digitalization influences job creation and destruction. Research Policy, 48(8), 103765. [CrossRef] Bartel, A., Ichniowski, C., & Shaw, K. (2007). How does information technology affect productivity? Plant-level comparisons of product innovation, process improvement, and worker skills. The Quarterly Journal of Economics,122(4), 1721–1758. [CrossRef] Blanchflower, D. G., & Burgess, S. M. (1999). New technology and jobs: Comparative evidence from a two country study. Economics of Innovation and New Technology,5(2–4), 109–138. Bris, A., & Cabolis, C. (2017). IMD world digital competitiveness ranking 2017; IMD Business School. Available online: https://investcroatia .gov.hr/wp-content/uploads/2018/01/world_digital_competitiveness_yearbook_2017.pdf (accessed on 10 February 2025). Brynjolfsson, E., & McAfee, A. (2014). The second machine age: Work, progress, and prosperity in a time of brilliant technologies. WW Norton & Company. Caravella, S., Cirillo, V., Crespi, F., Guarascio, D., & Menghini, M. (2023). The diffusion of digital skills across EU regions: Structural drivers and polarisation dynamics. Regional Studies, Regional Science,10(1), 820–844. Celeux, G., & Soromenho, G. (1996). An entropy criterion for assessing the number of clusters in a mixture model. Journal of Classification, 13(2), 195–212. [CrossRef] Chen, W., Niebel, T., & Saam, M. (2016). Are intangibles more productive in ICT-intensive industries? Evidence from EU countries. Telecommunications Policy,40(5), 471–484. [CrossRef] Cirillo, V., Fanti, L., Mina, A., & Ricci, A. (2023). The adoption of digital technologies: Investment, skills, work organisation. Structural Change and Economic Dynamics,67, 101–118. [CrossRef] Cohen, W. M., & Levinthal, D. A. (1989). Innovation and learning: The two faces of R&D. The Economic Journal,99(397), 569–596. Cohen, W. M., & Levinthal, D. A. (1990). Absorptive capacity: A new perspective on learning and innovation. Administrative Science Quarterly,35(1), 128–152. [CrossRef] Crisan, G.-A., Popescu, M. E., Militaru, E., & Cristescu, A. (2023). EU Diversity in Terms of Digitalization on the Labor Market in the Post-COVID-19 Context. Economies,11(12), 293. [CrossRef] Curzi, Y., & Ferrarini, F. (2024). High-performance work systems and firm innovation: The moderating role of digital technology and employee participation. Evidence from Europe. Management Research Review,47(13), 51–68. Damanpour, F., & Schneider, M. (2006). Phases of the adoption of innovation in organizations: Effects of environment, organization and top managers. British Journal of Management,17(3), 215–236. Della Torre, E., Gritti, A., & Salimi, M. (2021). Direct and indirect employee voice and firm innovation in small and medium firms. British Journal of Management,32(3), 760–778. [CrossRef] Draghi, M. (2024). The future of European competitiveness: A competitiveness strategy for Europe. European Commission. Available online: https://commission.europa.eu/document/download/97e481fd-2dc3-412d-be4c-f152a8232961 _ en?filename=The+future+ of+European+competitiveness+_+A+competitiveness+strategy+for+Europe.pdf (accessed on 10 February 2025). Eurofound & Cedefop. (2020). European Company Survey 2019: Workplace practices unlocking employee potential. European Company Survey 2019 Series. Publications Office of the European Union. Ford, M. (2015). The rise of the robots. Basic Books. García-Romanos, J., & Martínez-Ros, E. (2024). Digitalization and employment in Europe: The role of firm’s size and the complementarity of R&D. R&D Management,54, 695–712. Greenan, N., & Napolitano, S. (2024). Digital technologies, learning capacity of the organization and innovation: EU-wide empirical evidence from a combined dataset. Industrial and Corporate Change,33(3), 634–669. [CrossRef] Gregory, T., Salomons, A., & Zierahn, U. (2022). Racing with or against the machine? Evidence on the role of trade in Europe. Journal of the European Economic Association,20(2), 869–906. [CrossRef] Keynes, J. M. (1931). Economic possibilities for our grandchildren. In Essays in persuasion (pp. 358–374). Macmillan. Kongsted, A., & Nielsen, A. M. (2017). Latent class analysis in health research. Journal of Physiotherapy,63(1), 55–58. [CrossRef] Kozak, M., Kozak, S., Kozakova, A., & Martinak, D. (2020). Is fear of robots stealing jobs haunting European workers? A multilevel study of automation insecurity in the EU. IFAC Papersonline,53(2), 17493–17498. [CrossRef] Economies 2025,13, 104 19 of 19 Marx, K. (1867). Capital: A critique of political economy. Volume I: The process of capitalist production. Otto Meissner. Mastrostefano, V., & Pianta, M. (2009). Technology and jobs. Economics of Innovation and New Technology,18(8), 729–741. [CrossRef] Mitropoulos, P., Mitropoulos, A., & Vlami, A. (2024). Factors affecting high-quality entrepreneurial performance in smalland medium-sized family firms. Journal of Family Business Management,14(5), 974–996. Obadic, A. (2020). Influence of technological change and digital technology on job polarization and occupational change. In G. Druzic, & T. Gelo (Eds.), Conference proceedings of the international conference on the economics of decoupling, ICED 2020 (pp. 433–450). University of Zagreb, The Faculty of Economics & Business. Peng, G., & Zhang, D. (2020). Does information technology substitute for or complement human labor? A dynamic stratified analysis on European countries. Decision Sciences,51(3), 720–754. [CrossRef] Petersen, K. J., Qualter, P., & Humphrey, N. (2019). The application of latent class analysis for investigating population child mental health: A systematic review. Frontiers in Psychology,10, 1214. [CrossRef] Rogers, E. M. (1983). Diffusion of innovations (3rd ed.). Free Press. Santos, L. M., Amorim, L. D., Santos, D. N., & Barreto, M. L. (2015). Measuring the level of social support using latent class analysis. Social Science Research,50, 139–146. [CrossRef] Schumpeter, J. A. (1942). Capitalism, socialism, and democracy. Harper and Brothers. Siegenthaler, M., & Stucki, T. (2015). Dividing the pie: Firm-level determinants of the labor share. ILR Review,68(5), 1157–1194. Tornatzky, L. G., & Klein, K. J. (1982). Innovation characteristics and innovation adoption-implementation: A meta-analysis of findings. IEEE Transactions on Engineering Management,EM-29(1), 28–45. Van Oorschot, P., Hofman, E., Halman, J. I. M., & Bosch-Rekveldt, M. (2018). A bibliometric review of the innovation adoption literature. Technological Forecasting and Social Change,134, 1–21. Van Reenen, J. (1997). Employment and technological innovation: Evidence from U.K. manufacturing firms. Journal of Labor Economics, 15(2), 255–284. Wagenmakers, E.-J. (2007). A practical solution to the pervasive problems of p values. Psychonomic Bulletin and Review,14, 779–804. [CrossRef] [PubMed] Weller, B. E., Bowen, N. K., & Faubert, S. J. (2020). Latent class analysis: A guide to best practice. Journal of Black Psychology,46(4), 287–311. [CrossRef] World Bank. (2019). World development report 2019: The changing nature of work. World Bank. [CrossRef] Zhou, M., Thayer, W. M., & Bridges, J. F. P. (2018). Using latent class analysis to model preference heterogeneity in health: A systematic review. Pharmaco Economics,36, 175–187. [CrossRef] Zhu, L. (2024). Meritocratic beliefs in the United States, Finland, and China: A multidimensional approach using latent class analysis. The British Journal of Sociology,76(1), 153–172. [CrossRef] Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.