The Adoption of Digital Technologies - Investment, Skills, Work Organisation

Structural Change and Economic Dynamics 66 (2023) 89–105
Contents lists available at ScienceDirect
Structural Change and Economic Dynamics

journal homepage: www.elsevier.com/locate/strueco
The adoption of digital technologies: Investment, skills, work organisation

Valeria Cirillo a, b, *, Lucrezia Fanti e, Andrea Mina b, c, Andrea Ricci d
a
Department of Political Sciences, University of Bari ’Aldo Moro’, Bari, Italy
b
Institute of Economics and EMbeDS, Scuola Superiore Sant’Anna, Pisa, Italy
c
Centre for Business Research, University of Cambridge, Cambridge, United Kingdom
d
Istituto Nazionale per l’Analisi delle Politiche Pubbliche (INAPP), Rome, Italy
e
Department of Economic Policy, Università Cattolica del Sacro Cuore, Milan, Italy
A R T I C L E I N F O A B S T R A C T
JEL codes: New digital technologies are shaping the transformation of production activities. This process of change is
D20 characterised by growing digitization, inter-connectivity and automation. The diffusion of new technologies is,
L23 however, very uneven, and firms display different adoption behaviours. By using data on a representative sample
O33
of Italian firms, we explore the patterns and determinants of new digital technology adoption. We build our
Keywords: theoretical framework on the nexus between technology, skills and the organisation of work. We then provide
Digital technologies
novel econometric evidence on the positive effects of education and training. Among the notable results of the
Industry 4.0
Skills
paper, labour flexibility does not seem to favour new technology adoption, whereas second-level collective
Work organisation bargaining plays a positive role in the process. Results also show heterogeneous effects between large vs. small
and medium-size firms, and between manufacturing and service sectors.
1. Introduction may or may not lead to a radically new ‘techno-economic paradigm’

(Dosi, 1988; Freeman and Perez, 1988), there can be no doubt that the
Modern production technologies are characterised by increased high recombinatorial potential of digital technologies can foster new
digitization, automation and interconnectivity (Brynjolfsson and McA ways of performing economic activities. The ability of companies to
fee, 2014; Ford, 2015). These characteristics have been associated with select and exploit new sources of competitive advantage is, therefore, a
disruptive process innovations that have been qualified as new ‘enabling likely determinant of growth because of the expected
technologies’ (Teece, 2018). Martinelli et al. (2021) argue that these performance-enhancing attributes of these enabling technologies.
digital technologies can be defined as emergent technologies displaying In this contribution we focus on the technology adoption choices
some of the characteristics of general purpose technologies (GPTs) made by individual firms. While information on the production of new
(Bresnahan and Trajtenberg, 1995; Jovanovich and Rousseau, 2005; technologies is usually available on a large scale from standard firm-
Bresnahan, 2010), but are not yet fully developed GPTs and are deeply level data, or could be derived from the examination of patent re
dependent on the broader ICT paradigm where they function. In Europe cords, information on market diffusion is much rarer. This is a serious
these technologies are often subsumed in the policy debate under the shortcoming for research and for evidence-based policy because it is
umbrella term ‘Industry 4.0’. This concept, which has roots in German diffusion, rather than invention or innovation, the broad manifestation
industrial policy, captures the convergence of new operational tech of Schumpeterian structural change in the economy (Metcalfe, 1998;
nologies with Internet-driven IT, and marks a shift of production systems Stoneman and Battisti, 2010). Even though important work exists on the
towards the ‘smart factory’ of the future (Kagermann et al., 2013). diffusion of robots, for which the International Federation of Robotics
It is difficult to ascribe to a common matrix all the new digital provides aggregate data (Acemoglu and Restrepo, 2020; Dauth et al.,
technologies. These include a complex set of devices over a broad con 2017), the analyses of individual firm choices has proved much more
tinuum of production possibilities, and production choices are arguably difficult, as clearly pointed out by Seamans and Raj (2018). To
conditional on infrastructures and firm organization. While it remains to contribute to this important and under-researched aspect of the digital
be seen whether particular configurations of ‘enabling technologies’ economy, we explore a large and unique firm-level survey of Italian
* Corresponding author.
E-mail address: [email protected] (V. Cirillo).
https://doi.org/10.1016/j.strueco.2023.04.011
Received 12 July 2022; Received in revised form 14 April 2023; Accepted 15 April 2023
Available online 16 April 2023
0954-349X/© 2023 Elsevier B.V. All rights reserved.
V. Cirillo et al. Structural Change and Economic Dynamics 66 (2023) 89–105
Fig. 1. Percentage of firms investing in I4.0 technologies.

Note: sampling weights applied. Source: Authors’ elaborations on RIL 2015–2018 panel data
businesses: the survey “Rilevazione Imprese e Lavoro – RIL” (Longitu among Italian firms (section 2.2). Section 4 presents the empirical
dinal Survey of Businesses and Work), run by the Italian National strategy; while Section 5 contains the results of our econometric ana
Institute for the Analysis of Public Policies (INAPP). The survey contains lyses. Section 6 concludes by drawing attention on the strategic and
specific questions on different digital technologies acquired by firms, policy implications of the study, and by identifying new avenues for
thus offering a unique opportunity to study the determinants of adoption future research.
of new enabling technologies.
The specific focus of the paper is the nexus between technology, skills 2. Background literature and hypotheses
and the organisation of work. This is a fundamental aspect of the debate
about the disruptive potential of new digital technologies because the In the study of technical change, several factors have been identified
skills profile of firms filters the penetration of technical change in the as drivers of technology adoption choices, and much has been written on
economy and lay the foundations for the productivity gains that could be how these choices translate into aggregate patterns of technology
generated in the process (Acemoglu, 2002; Link and Siegel, 2003). After diffusion (for comprehensive reviews of this literature, see Karshenas
describing the aggregate patterns of adoption, we perform an econo and Stoneman, 1993; Hall and Khan, 2003; Stoneman and Battisti,
metric analysis of the role of skills, distinguishing between formal 2010).2 among the drivers of adoption, we can distinguish between: i)
human capital endowments and on-the-job training, and the role of two supply-side factors, including improvements of older technologies
essential aspects of the organisation of work within the firm: the use of leading to incremental innovations or changes in the use of existing
temporary contracts and the role of second-level (i.e. decentralised) technologies (Gruber and Verboven, 2001); ii) factors related to the
bargaining in the governance of the firm. Many empirical contributions demand for new technologies, such as technological complementarities
have highlighted the patterns of complementarity between high-skilled between producers and users, and the adopters’ stock of tangible and
workers and information technologies (among others, Autor et al., 2000; intangible capital (Rosenberg, 1976); iii) specific (internal) character
Bresnahan et al., 2002; and Fabiani et al., 2005; Piva et al., 2005, for istics of the firm, such as size,3 age, and human resource management
Italy). This body of empirical evidence seems to lend broad support to practices (Bloom et al., 2012); iv) sectoral heterogeneities between
the skill-bias technical hypothesis (Acemoglu, 2002). So far very few manufacturing and services (Comin and Mestieri, 2013; Calvino et al.,
studies have, however, been able to extend this line of enquiry to new 2018; Cirillo et al., 2022a); and finally v) the external organisation of the
digital technologies, and consider from a large-scale quantitative firm and the institutional context in which businesses operate (Dosi,
perspective the joint role of skills and work organisation in firm adop 1991; Mowery and Rosenberg, 1993). More broadly, the co-evolution of
tion.1 This study contributes to filling this gap in the literature.
The paper is organised as follows: in the first section we review the
literature on technology adoption and skills to frame our research 2
The literature on this topic is very rich, starting at least from the pioneering
questions and derive testable hypotheses (Section 2). In Section 3 we
contributions of Rogers (1962), Mansfield (1968), David (1969), Davies (1979),
present the data and provide some descriptive evidence on diffusion Metcalfe (1981) and Stoneman (1981), but its coverage is well beyond the
scope of this paper.
3
Firm size is indeed one of the main characteristics accounting for different
adoption patterns of digital technologies. As we show in Section 3 and Section
1
The few works that address this issue include Calvino et al. (2022) 4, firm size is particularly relevant in countries such as Italy, where SMEs
inspecting the role of workers’ skills as well as management capabilities, and represent the largest share of production infrastructures, leading to a slower
accumulation of intangible assets as main drivers of digital technologies in the aggregate innovation and technology adoption rate (Bugamelli et al., 2012;
Italian economy. ISTAT, 2017; ISTAT, 2018).
90
Fig. 2. Percentage of firms investing in I4.0 technologies.

organizational capabilities and economic environment shapes the (1994) distinguished between ‘specific human capital’, referring to skills
sources of competitive advantage (Nelson and Winter, 1982; Dosi et al., or knowledge that is useful only to a single employer, and ‘general
2000; Dosi and Marengo, 2015). Indeed, the sector and size dimensions human capital’ that is useful to many employers. In the case of general
can be linked to the concept of absorptive capacity (Cohen and Levin skills, theory suggests that the rewards for acquiring them accrue to the
thal, 1990) in explaining the patterns of technology adoption and employee in the form of higher wages, as determined by the market
diffusion. We speculate that factors such as skills, tasks, work organi (Prais, 1995). This reflects both the higher value of a skilled employee’s
zation at the core of our analysis are likely to accelerate the adoption contribution to the production process, and the cost (in time or money
processes when interacted with structural elements. Sectors are char terms) of training, which is shared by the employee. Specific skills, on
acterized by distinct trajectories, specific vertical linkages and techno the other hand, are only of value in a particular instance of employment,
logical regimes (Pavitt, 1984; Archibugi, 2001; Laursen and Meliciani, and are the result of ad-hoc training in a specific production context.
2002; Marsili and Verspagen, 2002; Malerba, 2002). Therefore, it is Innovations may also require very specific skill sets that do not yet
important to disentangle how and to which extent workforce and exist in the labour market. In this case workers have no direct incentive
organizational factors interact with demand conditions and structural to develop these skills because they are not immediately tradable, so the
factors. From this point of view, a growing literature has pointed to some cost of training falls on the employer. If, however, the employer is un
major specificities that make the innovation process in services mark able to fully appropriate the value of innovation and imitators can enter
edly different from that of manufacturing, and has emphasized the the market, skills that started as ‘specific’ can become more ‘general’,
increasing interdependence between the manufacturing and service and if these are only available in short supply, they will attract a high
branches of the economy (Evangelista, 2000; Guerrieri and Meliciani, market premium. Given the uncertainty surrounding the benefits of new
2005; Castellacci, 2008).4 technologies, and the heterogeneity of firm capabilities, calculations of
An important stream of contributions has specifically focused on the rates of return to skills will vary greatly and will contain large error
interplay between technology and human capital (OECD 2011), proxied margins.
by workers’ education levels as in standard human capital theory It is difficult to disentangle the skills that drive innovation from those
(Becker 1994) or conceptualised as workers’ knowledge and routines which are demanded as a result of change brought about by innovation
from an evolutionary perspective (Nelson and Winter, 1982; Winter, (Tether et al., 2005; Leiponen, 2005). It is nevertheless possible to ask
1997; Dosi and Marengo, 2015). Human capital can be accumulated which skill sets are more suitable for the adoption of new digital tech
through investments in education, training and any other means that nologies as part of firm-specific strategies. Firms have to anticipate skills
improve the employees’ ability to provide labour services. Becker needs in preparation for particular technology choices, and are bound in
their technology choices by path-dependent human capital endow
ments. Firms that chose to digitize all or part of their activities may
4
At least two groups of studies highlight the relevance of the sector dimen consider their position in the supply chain and identify an opportunity to
sion. On the one hand, there are those that focus on sector-specific techno achieve productivity gains. We argue that in making this strategic
logical regimes, and identify features that distinguish innovative activities and choice, firms recognise that these gains may only be realised if skilled
industrial dynamics across sectors by looking at technological opportunities, workers are able to use new technologies. In this sense, skilled workers
knowledge bases, cumulativeness and appropriability conditions (Marsili and foster the generation of productivity gains from the use of new tech
Verspagen, 2002; Malerba, 2002; Dosi et al., 2006). On the other hand, there nologies. Therefore, we propose the following baseline hypothesis:
are studies that look at patterns of industrial dynamics and emphasize the
relationship between sector-specific developmental paths and a variety of firm HP1. Firms with more skilled employees are more likely to invest in
innovation strategies (e.g. Pavitt, 1984). Building on this tradition, we expect new digital technologies.
heterogeneous findings when comparing adoption strategies in manufacturing
and in service industries. As we have already suggested, extensive theoretical and empirical
91
Table 1 plausible hypothesis that can be made for the new digital technologies
Percentage of firms investing in I4.0 tech by firm size and geographical area. follows the expectation that firms planning changes in production are
Firm size IoT Robotics Big Data Augmented Cybersecurity more likely to prepare their workers by updating or upgrading their skill
Analytics reality sets to fully capture the benefits of new technologies.6 Even though
5− 9 6.4 2.6 3.4 1.9 29.3 technical change always entails an element of uncertainty also in
10–49 7.6 5.9 6.7 2.7 34.9 diffusion processes, it can argue that firms expect technologies to require
50–249 14.0 14.1 14.3 4.7 53.4 contextual adjustments and may benefit from having training pro
> 250 28.1 21.8 27.4 9.4 68.2 grammes in place. For this reason, we posit that:
Total 7.7 5.1 5.9 2.5 33.9
Geographical HP2. The share of workers with on-the-job training has a positive ef
area
fect on the adoption of new digital technologies.
North West 8.3 5.8 5.6 3.4 38.9
North East 8.8 6.1 6.1 2.1 34.2 A related aspect of the decision to adopt new technologies is the
Centre 7.0 3.9 6.8 2.2 29.1
organisation of labour (Nelson and Winter, 1982; Osterman, 1994).
South 5.8 3.4 5.0 2.0 29.7
Total 7.7 5.1 5.9 2.5 33.9 Qualitative evidence on the processes of organisational adaptation in
dicates that this is as a necessary condition for the generation of pro
Note: sampling weights applied. Source: Authors’ elaborations on RIL
ductivity gains and competitive advantage stemming from the use of
2015–2018 panel data.
new enabling technologies (Fabbri et al., 2018). Process technologies,
such as ‘Industry 4.0’ technologies, tend to be accompanied by organ
literature indicates that the skills mix of firms is partly dependent on isational changes consistent with the principles of ‘lean production’
workers’ formal education and partly generated through specific in (Womack et al., 1990). This is an argument clearly made by Bresnahan
vestments in training. Several studies point to the role of high-skilled et al. (2002) in their seminal paper on the combination of computeri
and highly-trained human capital as key drivers of innovative activity zation, workplace organization and increased demand for skilled
and organizational change (see among others Leiponen, 2005; Lundvall, workers. Complementarity drove clusters of structural adjustments in
2009; Toner, 2011). High-quality human capital generates strong modern firms. More specifically, the use of information and communi
absorptive capacity (Cohen and Levinthal 1990) and is also associated cation technologies is positively correlated with increases in the demand
with the presence of new managerial practices (Böckerman et al., 2012) for various indicators of human capital and workforce skills. Moreover,
guiding the introduction of new capabilities and new organizational it can show patterns of correlations with specific forms of work orga
routines. While a firm’s absorptive capacity is more than the sum of the nization that include higher levels of labour flexibility and the use of
absorptive capacities of its employees (Cohen and Levinthal, 1990), it is short-term labour contracts.
related to the skills of employees, including those who stand at the On the one hand, labour flexibility can help firms to adjust quickly to
interface between the firm and its external environment. In case of rapid new technological requirements by allowing rapid changes to their de
and uncertain technical change, a deep knowledge base and/or some mand for labour (Bartelsman et al., 2016). This is often seen as an agile
predisposition for change might be required for effective communica way to optimize on labour endowments when production lines and
tion with technology experts or to capture, assimilate and exploit new whole firm sub-units are restructured to meet new market needs or in the
information for productive purposes.5 presence of increased market competition. On the other hand, evidence
Whereas general human capital can be acquired through the market exists that more flexible (internal) labour markets can hamper innova
process, more specific skills, tailored to the exact requirements of the tion at both firm and sector levels (see for example Kleinknecht et al.,
firm, can only be developed through training. As we expect skills to have 2014, Cetrulo et al., 2019, Hoxha and Kleinknecht, 2020). More reliance
a positive effect on adoption, we also expect training to favour the on temporary workers might lower the probability to invest in new
acquisition of new enabling technologies. The contribution of formal enabling technologies because it is more compatible with cost compet
education levels and training varies across firms and sectors, depending itiveness strategies rather than higher-value added models of techno
on the relative importance of ‘Occupational Labour Markets’ vs. ‘Inter logical advantages (Michie and Sheehan, 2003; Castro Silva and Lima,
nal Labour Market’ mechanisms of human capital upgrading (Rubery 2019). The accumulation of knowledge would thus be hampered by
and Grimshaw, 2003). The former depends on nationally recognized frequent employees turnover, preventing the attainment of productivity
qualifications; the latter is designed and organised as on-the-job training gains derived from idiosyncratic activities of learning by doing.7
by individual employers in accordance with their specific needs. All in all, we expect that continuous accumulation of tacit knowledge
Skills upgrading through training is beneficial to the adoption of new about production processes is important for the adoption of enabling
technologies because embodied technical change can only generate technologies and therefore we hypothesise that:
productivity gains it is successfully integrated in contextual production
processes (Boothby et al., 2010). Whenever skills requirements are HP3. The use of flexible staff arrangements has a negative effect on the
firm-specific, or in a context with stronger labour market frictions, adoption of new digital technologies.
training is the relative more efficient solution (Ramachandran, 1993; The final aspect to which we draw our attention is the role of firm
Hamermesh and Pfann, 1996). Moreover, in the complex cognitive
process of adaptation to radical new technologies (Raffaelli et al., 2019),
training can shape workers’ perception of a new technology and influ 6
Naturally, the two mechanisms can also coexist in a more ‘systemic inte
ence their attitude to change (Ouadahi, 2008). It is interesting to notice gration model’, for example combining higher-level science and engineering
that there are contrasting results in the literature on ICT adoption and skills of a small group of workers with highly trained employees. For the sake of
training: for example, Arvanitis (2005) found evidence of complemen clarity, however, it is important to stress that we take into account general
tary between investment in information technology and training, training, which may or may not be targeted to the usage of enabling or digital
whereas Giuri et al. (2008) did not. We would argue that the most technologies, the underlying idea being that firms characterized by a higher
emphasis on training are those presenting the more favourable organisational
environments for the adoption of new technologies.
7
It is also possible that flexible work is applied by firms to non-core tasks.
5
Along a similar line of research, McGuirk et al. (2015) have recently ana However, while this may be the case of large firms, it is a more unlikely
lysed the role of ‘Innovative Human Capital’, measured it as a combination of behaviour among small firms, which are less diversified and less complex or
education, training, but also willingness to change and job satisfaction, upon ganisations. We are going to subject this conjecture to empirical tests in the
small firms’ propensity to innovate. analysis of heterogeneous effects in Section 4.
92
governance in shaping technology adoption decisions. Companies that HP4. Second-level agreements have a positive effect on the adoption
plan to change or upgrade production technologies need to adapt ca of new digital technologies.
pabilities and skills as quickly as possible, and this may entail changes in
wages and the content of work. The structure of decision-making pro 3. Data and methodology
cesses within firms can therefore have significant effects on the outcome
of specific investment decisions. Company-level bargaining can be more 3.1. Context: digitalization in the Italian economy
flexible and versatile than centralized bargaining, thus offering some
advantages in more dynamic environments (Appelbaum and Berg, The secondary data available on the diffusion of digital technologies
1999). This is especially relevant when investment decisions concern among Italian firms show a scattered adoption of new enabling tech
new technologies. nologies. According to the Digital Economy and Society Index (DESI),
Lean production models associated with digital transformations which summarizes a set of indicators on Europe’s digital performance,
require a relatively high level of adaptation that might be negotiated Italy is placed at the bottom of the ranking in terms digital technologies
more easily at the company level rather than at the sectoral or national use (European Commission, 2018). This pattern is generally ascribed to
level. Company-level bargaining may cover specific topics such as well-known structural features of the Italian production system, above
workers’ involvement, changes in work organisation, working hours, all the large share of small and micro firms, and the share of value added
work roles, workloads, vocational training, and productivity premia.8 As coming from traditional sectors (Bugamelli et al., 2012). A recent report
already noted by Freeman and Medoff (1984) and Metcalf (2003), by the Italian Institute of Statistics (2018) highlighted the difficulty
economic theories have not been able to predict unambiguously the experienced by Italian companies in positioning themselves at the
impact of bargaining (and more generally unionisation) on firms. technological frontier and in exploiting the ongoing digital trans
Second-level (i.e. company-level) bargaining could both increase or formation through investments in technologies capable of reviving
decrease productivity. One might expect negative effects when the productivity dynamics. Among the factors hampering diffusion, there
conflictual behaviour of a trade union prevails; conversely, one might are stagnant macroeconomic growth dynamics, strong territorial
expect a positive impact if workplace unionism and collective bargai dualism, the preeminent weight of small and medium-sized enterprises
ning are set in a localised collaborative and participatory environment. on national production and a low average propensity to innovate, with
Several studies contain empirical investigations of the link between negative outcomes for both productivity and employment (Codogno,
company-level agreements and firm performance (Frick and Möller, 2009; Calligaris et al., 2016; ISTAT, 2017).
2003; Fairris and Askenazy, 2010; Jirjahn and Mueller, 2014; Devicienti The data provided by the survey on the use of ICTs also run by the
et al., 2017; Antonietti et al., 2017; Damiani et al., 2018; Garnero et al., Italian Statistical Office (ISTAT) contain interesting contextual infor
2019), but the evidence is contradictory and overall inconclusive. mation on the state-of-the-art of digitization in the country. ISTAT
Divergent results have also been found by several studies of the relation (2018) reports a significant increase in the number of production units
between collective bargaining and innovation (see Menezes-Filho and that have introduced the use of ITC technologies to support business
Van Reenen, 2003; Addison and Wagner, 1997; FitzRoy and Kraft, 1990; data sharing (ERP) in various sectors (approximately 36.5% in 2017,
Schnabel and Wagner, 1994). compared to 21.5% in 2012), with particular emphasis on the automo
Kleinknecht (2020) argues that decentralised bargaining would tive and telecommunications sectors. Against this backdrop, more
hamper innovation because it can induce firms to use downward wage detailed microeconomic analyses are much needed to acquire deeper
flexibility rather that innovation or technology adoption to remain understanding of the dynamics at play.
competitive. In a rare contribution specifically focused on digital tech
nologies and based on establishment survey data and employee data, 3.2. Data
Genz et al. (2019) have recently found a strong negative relation be
tween work councils and investments in Industry 4.0 technologies in Our empirical analysis is based on an original database drawn from
Germany. The implementation of digital technologies broadens the re the ‘Rilevazione Imprese e Lavoro’ (RIL) survey conducted by INAPP in
sponsibility of work councils to mediate the conflict between employees 2015 and 2018 on a representative sample of companies (including both
and management. Work councils can exert veto rights with respect to the listed, non-listed companies and limited liability firms). Each wave of
implementation of digital technologies and can narrow the freedom of the survey covers over 30,000 firms operating in non-agricultural pri
action of management. They tend to support the implementation of vate sectors. A sub-sample of the firms included in the survey (around
digital technologies only in those establishments characterised by a high 45%) are followed over time, making the RIL dataset a partial panel over
share of workers performing physically demanding jobs or subject to the period under investigation.9
competitive pressures (Genz et al., 2019). The RIL-INAPP survey collects a rich set of information about the
Building on a qualitative research approach and in-depth interviews composition of the workforce, including the amount of investments in
with trade unions’ delegates and managers of Italian companies, Cirillo training, hiring and separations, the use of flexible contractual ar
et al. (2021) detect instead a lack of trade unions’ involvement in the rangements, the asset of the industrial relations and other workplace
design phase of Industry 4.0 artefacts, regardless of the degree of digi characteristics. Moreover, the data contains an extensive set of firm level
talisation and robotisation in action. However, the authors also suggest controls, including management and corporate governance character
that trade unions play a key role in the implementation of new tech istics, productive specialization and other variables proxying firms’
nologies by encouraging workers’ acceptance and adaptation. strategies (such as propensity to introduce product and process in
Second-level bargaining could therefore allow firms to better appro novations and share of export on value added).
priate the gains of technological and organisational improvements. The fifth wave of the RIL-INAPP survey collected information on the
Through the implementation of second-level bargaining, trade unions introduction of Industry 4.0 technologies – hereafter I4.0 technologies.
might also be able to foster a collaborative environment and create A specific question was added on investments in new technologies over
preconditions for work practices that could improve motivation, job the period 2015–2017: “In the period 2015–2017 did the firm invest in
quality and productivity (Huselid, 1995). In light of this qualitative
evidence, our fourth and final hypothesis is that:
9
The RIL Survey sample is stratified by size, sector, geographical area and the
legal form of firms. Inclusion depends on firm size, measured by the total
8
Content and types of second-level agreements vary widely across countries, number of employees. For more details on RIL questionnaire, sample design and
sectors and firms (see Kleinknecht, 2020, for a discussion of this issue). methodological issues see: http://www.inapp.org/it/ril.
93
Table 2
Percentage of firms investing in I4.0 technologies by sector.
At least one IoT Robotics Big Data Analytics Augmented Cybersecurity
reality
Mechanical industry 57.2 12.5 14.7 7.0 3.9 46.9

Financial services 54.9 7.4 0.7 11.0 2.1 51.9
ICT 48.6 11.8 2.1 15.7 7.5 43.2
Chemical industry 45.2 9.7 17.3 7.2 4.0 36.8
Education, Healthy 43.1 11.1 3.3 7.9 1.1 37.6
Other manufacturing 42.4 7.8 7.8 4.1 1.7 36.1
Other services 41.8 9.3 2.1 6.2 4.7 37.5
Mining 40.2 7.7 2.6 3.5 1.2 36.2
Food industry 38.7 6.7 8.6 5.3 1.7 33.1
Trade 38.1 7.5 1.8 6.8 2.1 35.0
Textile industry 38.1 7.2 6.2 2.4 1.9 32.1
Transport 34.3 5.4 5.3 8.8 0.9 26.6
Constructions 33.8 2.9 0.8 2.0 1.2 31.3
Hotel and restaurants 21.0 5.5 0.7 2.5 1.6 16.2
Total 39.1 7.7 5.1 5.9 2.5 33.9
Note: sampling weights applied. Source: Authors’ elaborations on RIL 2015–2018 panel data.
Fig. 3. Education, training, temporary work by I4.0 investors.

these new technologies?”. The respondent was presented with the excluding observations with missing values for the key variables used in
following options: Internet of things (IoT), Robotics, Big data analytics, the analysis, the final sample is given by a panel of around 8000 firms
Augmented reality and Cybersecurity. It was possible to give multiple observed in each year considered.
answers. The timing of this survey is important: the data were collected
right after the implementation of the ‘National Enterprise Plan 4.0’, an
3.3. Descriptive evidence
incentives scheme that was specifically designed by the Italian Gov
ernment to lower financial constraints to investment and accelerate the
Fig. 1 shows uneven patterns of adoption, with less than 40% of
diffusion of I4.0 technologies. All firms were eligible to the scheme and
companies with at least 5 employees stating that they had made at least
all of them automatically received the incentive if they invested.10 In
one investment in I4.0 technologies over the period 2015–2017
what follows we present descriptive evidence on the diffusion of tech
(Fig. 1).11
nologies, disaggregating the data by sector, firm size and geographical
A relatively small group of firms shows combined adoption of several
location.
I4.0 technologies. Almost 30% of firms invest only in one type of
Our empirical analysis is performed on firms with at least 5 em
ployees in order to examine only productive units with a minimum level
of organisational structure. After imposing this selection criterion and 11
This percentage falls to 26% if we consider all the 30.000 firms (i.e. all firms
with at least one employee) interviewed in 2018.
10
This means that there was no selection or self-selection into the scheme.
94
Fig. 4. Share of firms signing second-level bargaining by number of I4.0 technologies.

technology – showing a ‘single-technology’ approach to digitalization.12 regions, which are well-connected with international value chains,
These percentages fall when we consider simultaneous investments in invest more in new technologies than their counterparts located in
more than one I4.0 technology: only 7.4% of the RIL-INAPP panel Central and Southern regions. Cybersecurity emerges as the most
invested in two types of technologies, 2.7% invested in three technolo frequently adopted new digital technology also from analyses compiled
gies, 0.5% in four technologies, 0.3% in all I4.0 technologies (IoT, Ro by the Italian National Institute of Statistics (ISTAT, 2021). By relying on
botics, Big Data Analytics, Augmented reality and Cybersecurity). All in the “Permanent Census on Enterprises”, ISTAT shows that in the period
all, the diffusion of the new Industry 4.0 paradigm is generally limited to 2016–2018 about 27% of Italian firms with at least 10 employees
a ‘single technology’ adoption approach rather than a ‘multi-technol invested in cybersecurity technologies, and this percentage rises to
ogy’ strategy based on simultaneous investments in complementary about 65% when we look at large firms with more than 250 employees
technologies. Unsurprisingly, the adoption of I4.0 techs is associated (ISTAT, 2021).13 Again, firm size plays a prominent role, as confirmed
with firm size, highlighting the existence of well-known differences in by our data. Moreover, companies located in the North East and North
cost barriers between small and large and varying absorptive capacities West regions, which are well-connected with international value chains,
(Cohen and Levinthal, 1990). invest more in new technologies than their counterparts located in
The percentage of firms adopting at least one I4.0 technology in Central and Southern regions.
creases with firm size (see Fig. 2) – from 33% in small firms (5–9 em Table 2 illustrates the distribution of I4.0 adopters across sectors.
ployees) to 77% in large enterprises (more than 250 employees). A The sectors featuring a greater incidence of companies investing in the
‘multi-technology’ approach to I4.0 adoption is also strongly associated new enabling technologies are chemistry (45.2%), financial services
with firm size: the percentage of firms introducing contemporaneously (54.9%), information and communication (48.6%) and mechanics
five I4.0 technologies increases from 0.3% of micro firms (5–9 em (57.2%). ICT displayed a higher incidence of I4.0 adopters in IoT
ployees) to 1.2% among large business (i.e. firms with more than 250 (11.8%), Big Data (15.7%) and Cybersecurity (43.2%). The mechanical
employees). and chemical sectors are instead characterised by the most pronounced
Tables 1 and 2 show the distribution of I4.0 adopters by type of adoption of robotics.
technology, firm size and firm location (Table 1) and by sector (Table 2). Linking the investment in new digital technologies to the skill mix of
A first distinction can be made between companies that invested in ‘at firms, Fig. 3 shows that on average firms that report investments in the
least one’ enabling technology and companies that introduced a specific new technologies have a higher share of tertiary educated workers and
typology among those indicated, i.e., ‘IoT’, ‘Robotics’, ‘Big Data’, register a more intense use of training activities. Conversely, I4.0
‘Augmented Reality’ and ‘Cybersecurity’. Table 1 shows how the diffu adopters register on average a lower share of employees on short-term
sion of new enabling technologies is strongly influenced by the size of contracts. There may be several reasons for this, including lack of in
the enterprises. Cybersecurity is the most frequent choice: on average formation on the returns to education and training, lack of access to
33.9% of Italian firms invested in some forms of information security, finance for training, or even fear of not reaping the returns on invest
while augmented reality and robotics concern a smaller share of Italian ment in training because of the risk of poaching. The latter might be
companies – respectively 2.5% and 5.1%. Again, firm size plays a especially relevant to highly-innovative and high-tech industries.
prominent role. Companies located in the North East and North West When it comes to industrial relations, we detect a positive
12 13
This percentage falls to 20% including firms with less than 5 and more than The census covers 2,80,000 enterprises located in Italy with at least 3
1 employees, and to 15% if we include companies without employees. workers (for further details see https://www.istat.it/it/archivio/238337).
95
Table 3
Estimates of the average marginal effects. Dep var: Probability to invest in I4.0 and number of I4.0 technologies.
(1) (2) (3) (4)
At least one Number of I4.0 At least one Number of I4.0
b/se b/se b/se b/se
Share of college workers 0.177028*** 0.396379** 0.115041*** 0.266226**

Share of trained workers (0.037) (0.155) (0.035) (0.111)
0.043859*** 0.087007*** 0.060826*** 0.003022
Share of fixed-term workers (0.013) (0.026) (0.012) (0.040)
− 0.063911* 0.020456 − 0.004203 − 0.022740
Second level agreement (0.036) (0.100) (0.035) (0.104)
0.036378** 0.117391*** 0.037491*** 0.066634
(0.017) (0.041) (0.014) (0.043)
Firm size 0.070322*** 0.002281*** 0.074574*** 0.000298***
VA per worker (log) (0.005) (0.001) (0.005) (0.000)
0.009902** 0.034536*** 0.012234*** 0.028623**
Share of workers over 50 (0.005) (0.009) 0.004 (0.013)
− 0.050792 − 0.301616*** − 0.063056** − 0.333603***
Share of workers 35–50 (0.033) (0.068) (0.030) (0.095)
0.029385 − 0.036414 − 0.000807 − 0.005286
Family firm (0.029) (0.057) (0.028) (0.088)
− 0.000216 − 0.083669*** − 0.022762* − 0.126162***
Share of workers with (0.015) (0.028) (0.013) (0.041)
high school degree 0.096306*** 0.079446* 0.051784*** 0.062981
In a trade group (0.021) (0.047) (0.019) (0.063)
0.032560*** 0.076495*** 0.044883*** 0.035137
Management with (0.012) (0.025) (0.011) (0.036)
tertiary education 0.029777* 0.108422*** 0.012871 0.059479
Management with (0.018) (0.039) (0.016) (0.048)
high school degree 0.009442 0.059701* − 0.004793 0.051265
Female management (0.015) (0.035) (0.014) (0.041)
0.012188 − 0.014869 0.023233 0.002512
Product innovation (0.017) (0.033) (0.015) (0.045)
0.065874*** 0.159434*** 0.088955*** 0.106009***
Process innovation (0.013) (0.025) (0.011) (0.036)
0.081530*** 0.186082*** 0.123046*** 0.060461*
Firm age (0.013) (0.025) (0.012) (0.036)
0.000136 0.000540** 0.000233* 0.000924*
FDI (0.000) (0.000) (0.000) (0.000)
0.090194*** 0.207316*** 0.104813*** 0.287045***
Share of export on VA (0.028) (0.060) (0.023) (0.074)
0.000168 0.000536 0.000246 0.001423*
Sec. and reg. dummies (0.000) (0.000) (0.000) (0.001)
Yes Yes Yes Yes
Observations 7746 7746 7675 7675

Non zero obs. 3719 3413 3.413
Censored obs
Uncensored obs 996.03 714.49 4262 4262
Wald Chi2 1479.55 375.60
Prob > Chi2 0.0000 0.0000 0.0004 0.0000
Pseudo R2 0.1051
Sample sel. stat.: 1,7084*** − 0.4169
athrho
LR test (rho = 0): 0.4801 0.0535
12.66 60.66
chi2(1) = 0.0000 0.0000
Prob > Chi2 0.0004 0.0000
1 number of employees in columns (2) and (4), log of number of employees in (1) and (2).
Note: marginal effects (1); Zero-Inflated Poisson (2); Heckprobit selection model (3); Heckman selection model (4). Omitted variables: managers with lower secondary
and primary education, workers with lower secondary and primary education; workers less than 35 years old. First stage exclusion restrictions "financially weak" (3, 4):
“During 2014 did the company apply for a bank credit for cash or liquidity reasons?”.
***
p<0.01,.
**
p<0.05,.
*
p<0.1.
Source: Authors’ elaborations on RIL 2015–2018 panel data.
96
Table 4 Table A1
Estimates of the average marginal effects by firm size and sector of activity. Dep Descriptive statistics.
var: Probability to invest in I4.0 and number of I4.0 technologies. Mean Sd Min Max
(1) (2) (3) (4)
At least one 0.39 0.4879759 0 1
At least one At least one At least one At least one
Number of I4.0 tech 0.5511074 0.835768 0 5
< 250a > 250 Manuf. Serv.
IoT 0.0773866 0.2672184 0 1
b/se b/se b/se b/se
Robotics 0.0506843 0.2193641 0 1
Share of college 0.175490*** 0.254780** 0.271261*** 0.157283*** Augmented Reality 0.0254782 0.1575808 0 1
workers Big Data Analytics 0.0588266 0.2353126 0 1
Share of trained (0.039) (0.120) (0.082) (0.046) Cybersecurity 0.3387317 0.4733041 0 1
workers 0.045031*** 0.024989 0.037894* 0.038673* Share of college workers 0.0950232 0.185736 0 1
Share of fixed- (0.014) (0.050) (0.020) (0.021) Share of trained workers 0.3446457 0.4181189 0 1
term workers − 0.058052 − 0.180278* − 0.013164 − 0.036111 Share of fixed-term workers 0.1062658 0.1998845 0 1
Second level (0.038) (0.108) (0.064) (0.049) Second level agreement 0.0615791 0.2404027 0 1
agreement 0.044560** − 0.091568** 0.048062** − 0.008001 Firm size 28.83787 184.4305 5 10,852
Firm size (0.019) (0.042) (0.024) (0.029) VA per worker (log) 11.70945 1.185182 0.089 16.106
0.065919*** 0.112861*** 0.087764*** 0.061900*** Share of workers over 50 0.2210545 0.2198158 0 1
VA per worker (0.007) (0.022) (0.009) (0.008) Share of workers 35–50 0.4611166 0.2590163 0 1
(log) 0.008292* 0.016044 0.009562 0.008513 Family firm 0.9073621 0.2899402 0 1
Share of (0.005) (0.011) (0.007) (0.007) Share of workers with high school 0.497609 0.3186729 0 1
workers over − 0.054325 0.036595 0.011375 − 0.078847 degree
50 In a trade group 0.5717729 0.4948486 0 1
Share of (0.034) (0.128) (0.049) (0.051) Management with tertiary 0.2322171 0.4222696 0 1
workers 0.031214 − 0.042711 0.043077 0.018206 education
35–50 Management with high school 0.5614778 0.496233 0 1
Family firm (0.030) (0.118) (0.045) (0.043) degree
0.000018 − 0.012948 0.044743** − 0.053089** Female management 0.1612228 0.3677563 0 1
Share of (0.016) (0.034) (0.022) (0.022) Product innovation 0.3594212 0.4798568 0 1
workers with 0.098773*** 0.042262 0.059836* 0.120764*** Process innovation 0.2855514 0.4517011 0 1
high school Firm age 25.7691 21.09899 0 1009
degree FDI 0.0172501 0.130209 0 1
In a trade group (0.022) (0.086) (0.032) (0.033) Share of export on VA 7.191898 19.23426 0 100
0.030456** 0.058112 0.041175** 0.031277* Financially weak 0.2744067 0.4462397 0 1
Management (0.013) (0.045) (0.018) (0.018)
Note: sampling weights applied.
with tertiary 0.032018* − 0.024698 0.034329 0.024055
education Source: Authors’ elaborations on RIL 2015–2018 panel data.
Management (0.018) (0.075) (0.025) (0.030)
with high 0.009351 0.005985 0.031547 − 0.007562 businesses and record a relatively low incidence of second-level
school degree
bargaining.14
Female (0.015) (0.073) (0.021) (0.027)
management 0.012480 0.037783 − 0.005873 0.030459 Having sketched the broad patterns of diffusion, we now move on to
Product (0.018) (0.067) (0.026) (0.025) investigate the determinants of firm technology choices in a multivariate
innovation 0.068405*** − 0.003449 0.038097** 0.100257*** setting.
Process (0.014) (0.044) (0.019) (0.020)
innovation 0.085311*** 0.038322 0.101617*** 0.034341
Firm age (0.014) (0.047) (0.018) (0.022)
3.4. Empirical strategy
0.000138 0.000315 − 0.000016 0.000364
FDI (0.000) (0.000) (0.000) (0.000)
0.079708** 0.066113 0.051836 0.103243* The central research questions of this paper focus on the role of skills,
Share of export (0.033) (0.044) (0.036) (0.053) training and the organisation of work as determinants of technology
on VA 0.000265 − 0.001320* − 0.000137 0.000603
adoption. We estimate the following equation:
Sec. and reg. (0.000) (0.001) (0.000) (0.001)
dummies Yes Yes Yes Yes Yi,t = α + β1 Ei,t− 1 + β2 Ti,t− 1 + β3 FTi,t− 1 + β4 SBi,t− 1 + β5 Xi,t− 1 + ui,t t
Observations 7206 536 3533 3233 = [2015, 2018] (1)
Wald Chi2 773.15 105.02 519.69 345.66
Prob > Chi2 0.0000 0.0000 0.0000 0.0000 where the dependent variable of Eq. (1) (Yi,t ) represents, alternatively:
Pseudo R2 0.0851 0.1836 0.1219 0.0852
(i) a dichotomous indicator (I4.0) taking value 1 if the firm i has invested
Note: omitted variables: managers with lower secondary and primary education, in at least one I4.0 technology over the period 2015–2017, and
workers with lower secondary and primary education; workers less than 35 0 otherwise; (ii) a categorical indicator (Number I4.0) assuming discrete
years old. values from 0 to 5 according to the total number of I4.0 technologies
***
p<0.01. introduced. As for our key explanatory variables, Ei, Ti, FTi and SBi
**
p<0.05,.
* indicate, respectively, the share of workers with college education, the
p<0.1.
share of trained workers, the share of fixed-term workers, that is a proxy
Source: Authors’ elaborations on RIL 2015–2018 panel data.
a
Firms included are those having at least 5 employees and less than 250 for flexible within-firm work arrangements (measured in 2015), and the
employees. adoption of second-level bargaining. Analogously, vector Xi includes a
set of lagged controls for management and corporate governance,
workforce composition, firms’ productive and competitive characteris
tics as well as industrial relations (see Table A1 in appendix for full
association between second-level agreements and the number of new descriptive statistics of the control variables), while the parameter ui,t
digital technologies introduced (see Fig. 4). On average, firms recurring
to decentralized bargaining show a higher incidence of new technology
adoption with respect to those relying on other levels of bargaining. 14
This is in line with the suggestion that larger firms could face more resis
Leading multiple-technology adopters – those reporting simultaneous tance by unions due to concerns that the new technologies may exacerbate the
investments in all types of digital technologies – tend to be large risk of control over workers (Moro et al., 2019).
97
Table A2 suffer from potential omitted variable bias and reverse causality.15 The
Test of exclusion restriction. inclusion of a broad set of controls allows us to minimize the endoge
(1) (2) neity bias arising from the presence of omitted variables. Moreover,
Investment I4.0 Number of I4.0 including these variables as pre-determined controls helps to address
b/se b/se reverse causality concerns.16
Share of college workers 0.182506*** 0.297719*** We also implement a two stage Heckman procedure (Amemiya 1985;
Share of trained workers (0.037) (0.074) Heckman 1979) conditioning the adoption choice on the likelihood that
0.042909*** 0.089882*** firms were investment-active. In our framework this is equivalent to
Share of fixed-term workers (0.013) (0.025)
− 0.063130* − 0.010188
performing a binary or count response model with sample selection
Second level agreement (0.037) (0.062) (Wooldridge 2010), as follows:
0.037520** 0.162264*** ( )
Firm size (0.017) (0.034) Pr Ii,t− 1 = α + β1 Ei,t− 1 + β2 Ti,t− 1 + β3 FTi,t− 1 + β4 SBi,t− 1 + β5 Xi,t− 1
0.070131*** 0.000369*** + γZi,t− 1 + ui,t (2)
VA per worker (log) (0.005) (0.000)
0.009296** 0.031605***
Share of workers over 50 (0.005) (0.009) Yi,t∗ = α + β1 Ei,t− 1 + β2 Ti,t− 1 + β3 FTi,t− 1 + β4 SBi,t− 1 + β5 Xi,t− 1 + λi + εi,t
− 0.050115 − 0.257149*** (3)
Share of workers 35–50 (0.033) (0.057)
0.031327 − 0.028903 The dependent variable Pr(Ii,t-1) in the selection Eq. (2), is a proba
Family firm (0.029) (0.052) bility index assuming value 1 if firm i made investments in 2015 and
0.000794 − 0.128519*** 0 otherwise. As for the explanatory variables, vector X includes the
Share of workers with high school degree (0.015) (0.031)
0.096769*** 0.088260**
entire set of controls already considered in Eq. (1). We use as exclusion
In a trade group (0.021) (0.038) restriction a variable indicating the firm’s demand for bank loans due to
0.033435*** 0.069015*** cash or liquidity problems11. The dependent variable Y*i,t in Eq. (3),
Management with tertiary education (0.012) (0.022) represents either the dummy indicator for any I4.0 technology, or the
0.030788* 0.103339***
number of I4.0 technologies (Number I4.0). This is observed mainly if Pr
Management with high school degree (0.018) (0.032)
0.009667 0.051493** (It-1) is equal to 1, i.e. if firm i realised an investment in 2015. Accord
Female management (0.015) (0.026) ingly, the right-hand side variables are the same set of controls for
0.012318 − 0.018181 managers, firms and workforce characteristics use in Eq. (1), while λi is
Product innovation (0.017) (0.030) the Inverse Mills Ratio (IMR) accounting for self-selection into the in
0.066726*** 0.162801***
Process innovation (0.013) (0.025)
vestment decision12. The Probit equation for the probability of firm in
0.081139*** 0.207712*** vestment in 2015 is therefore completely observed on data, while the
Firm age (0.013) (0.026) selected sample is available for analysing the impact of human capital
0.000135 0.000732** mix on I4.0 technology adoption. Eqs. (2) and (3) are estimated simul
FDI (0.000) (0.000)
taneously by Maximum Likelihood. This procedure allows us to correct
0.089610*** 0.407903***
Share of export on VA (0.028) (0.067) for sample selection bias and to obtain consistent estimates of average
0.000173 0.001701*** marginal effects. Auxiliary information and the main statistical tests for
Financially weak (0.000) (0.001) the sample selection hypothesis are reported in the last rows of Table 3
− 0.005122 0.003832 and 4, while results from estimation of Eq. (2) are reported in Table A2
Constant (0.012) (0.022)
− 0.182672
of Appendix.
(0.159)
Observations 7707 7707 4. Results
Wald Chi2 991.71 24.05
F(52, 7654) 0.0000
Prob > Chi2 0.0000 Table 3 contains the main results of our econometric analysis. First of
Prob > F 0.0000 all, this concerns the relationship between skills and i) the probability of
Pseudo R2 0.1052 0.1737 adoption (first column) and ii) the intensity of adoption measured by the
R2
number of technologies (second column). We find that firms charac
Root MSE 0.88051
terized by a higher share of educated workforce invest more in new
Note: marginal effects (1); Zero-Inflated Poisson (2). Omitted variables: man technologies and are more likely to display a multi-technology adoption
agers with lower secondary and primary education, workers with lower sec strategy. Ten percentage points increase in the share of tertiary educated
ondary and primary education; workers less than 35 years old. workers are associated with a raise of 1.8% in the probability of intro
***
p<0.01,.
** ducing at least one I4.0 technology and 3.9% increase in the number of
p<0.05,.
* technologies introduced. Similarly, a higher proportion of trained
p<0.1.
Source: Authors’ elaborations on RIL 2015–2018 panel data. workers also appears to favour both the probability of investment and
the number of digital technologies adopted - a ten percentage points
increase in the share of trained workers is correlated with 0.4% change
indicates an idiosyncratic error term.
in the probability of I4.0 and 0.9% increase in the number of digital
Non-linear regression models are used to estimate different specifi
cations of Eq. (1). We run a Probit and a Zero Inflated Poisson model to
estimate, respectively, the average marginal effects associated to the 15
This happens, for example, when implicit social norms at workplace and
probability of introducing at least one technology and the total number
managers’ personal traits, typically not observed by the researcher, affect both
of new digital technologies (Wooldridge, 2010; Cameron and Trivedi, the quality of human resource practices (low workers turnover, high share of
2013). Potential issues concerning unobserved heterogeneity and sam skilled and trained workers, and so on) as well as firms’ innovative and pro
ple selection (endogeneity) may arise in this set-up. Namely, if there are ductive behaviour. In this case, a positive non-linear estimate in Eq. (1) may
both observable and unobservable factors simultaneously affecting reflect firms and managers’ unobserved characteristics rather than the impact
workforce human capital endowment and the propensity to invest in of human capital on the adoption of I4.0 technologies.
16
new technologies, the Probit and Zero Inflated Poisson estimates might In particular, we refer to variables related to corporate governance, de
mographic profile of managers, recruitment policies and industrial relations.
98
Table A3
Robustness check: alternative measures for training.
(1) (2) (3) (4)
b/se b/se b/se b/se
Share of college workers 0.177591*** 0.382831** 0.174657*** 0.449092***

(0.037) (0.157) (0.041) (0.174)
Cost of training entirely borne by the company 0.033392*** 0.035251*
(0.011) (0.021)
Log cost of training per employee 0.010313*** 0.022758***
(0.002) (0.005)
Share of fixed-term workers − 0.067809* 0.017638 − 0.060195 0.051357
(0.037) (0.094) (0.039) (0.098)
Second level agreement 0.039705** 0.118171*** 0.038563** 0.109746**
(0.017) (0.038) (0.018) (0.043)
Firm size 0.071687*** 0.002316*** 0.065056*** 0.001872***
(0.005) (0.001) (0.006) (0.001)
VA per worker (log) 0.010046** 0.032698*** 0.008049 0.028552***
(0.005) (0.010) (0.005) (0.010)
Share of workers over 50 − 0.049340 − 0.301959*** − 0.044252 − 0.278547***
(0.033) (0.067) (0.035) (0.076)
Share of workers 35–55 0.029032 − 0.033104 0.035668 − 0.016383
(0.029) (0.057) (0.031) (0.062)
Family firm − 0.000831 − 0.080314*** 0.001509 − 0.080648***
(0.015) (0.028) (0.016) (0.031)
Share of workers with high school degree 0.097336*** 0.081078* 0.097144*** 0.068179
(0.021) (0.047) (0.022) (0.050)
In a trade group 0.034696*** 0.080319*** 0.036328*** 0.088475***
(0.012) (0.025) (0.013) (0.027)
Management with tertiary education 0.032218* 0.111312*** 0.025698 0.102637**
(0.018) (0.038) (0.019) (0.041)
Management with high school degree 0.011090 0.061231* 0.007745 0.057735
(0.015) (0.034) (0.016) (0.037)
Female management 0.014156 − 0.011896 0.011474 − 0.007646
(0.017) (0.033) (0.018) (0.036)
Product innovation 0.067437*** 0.163005*** 0.061172*** 0.143279***
(0.013) (0.025) (0.014) (0.027)
Process innovation 0.082570*** 0.188848*** 0.075333*** 0.182318***
(0.013) (0.025) (0.014) (0.027)
Firm age 0.000144 0.000563** 0.000171 0.000506**
(0.000) (0.000) (0.000) (0.000)
FDI 0.092977*** 0.202763*** 0.083316*** 0.219513**
(0.028) (0.050) (0.030) (0.094)
Share of export on VA 0.000157 0.000501 0.000401 0.000898**
(0.000) (0.000) (0.000) (0.000)
Number of employees 0.002330*** 0.001899***
(0.001) (0.001)
Observations 7746 7746 6774 6774

Non zero obs. 3719 3225
Wald Chi2 992.61 708.19 905.26 663.91
Prob > Chi2 0.0000 0.0000 0.0000 0.0000
Pseudo R2 0.1050 0.1091
technologies. Conversely, the share of fixed-term workers is negatively the probability of adoption and 6.6% increase for the number of new
associated to the adoption choice, even though this result is statistically technologies. This means that controlling for firm propensity to invest,
weaker, but is not correlated with the adoption of multiple technologies. these enabling technologies remain strongly associated with skills and
More in detail, a ten percentage points increase in fixed-term employees training as well as to “cooperative” industrial relations at the workplace
decreases the probability of I4.0 by 0.6%. Moreover, second-level level.17 As we mentioned in the discussion of the descriptive results
agreements seem to favour adoption and exert a positive effect on (section 2.2), the probability of introducing new digital technologies
both the adoption of technologies (by 3.6%) and the number of tech and the adoption of a multi-technology strategy are strongly and posi
nologies (by 11%). tively correlated with a wide set of control variables. These provide very
When we control for selection bias (columns 3 and 4), the share of useful insights into complementary (firm) characteristics that may
fixed-term workers and second-level agreements (for the number of trigger investment in new enabling technologies. The results suggest
technologies) are no longer significant, but both skills and training that larger firms, led by higher quality management in terms of educa
maintain a positive and significant estimated correlations – even though tional level, higher productivity levels, more innovative (both in terms
lower in magnitude - on the probability to invest in digital technologies. of process and product innovation), and more internationalised, present
Specifically, ten points increase in the share of tertiary educated workers
leads to 1.1% raise of the probability of introducing I4.0 technologies
and 2.6% raise of the probability of number of digital technologies. 17
Table A3 in the Appendix shows a robustness check performed by replacing
Similarly, a higher share of trained workers increases the probability of the variable ‘share of trained workers’ with a continuous variable proxying the
introducing I4.0 by 0.6%. Considering the potential selection issue, we average cost in training per employee and alternatively a dichotomous variable
also confirm the positive association between the presence of second- taking value of one if the cost of training is fully borne by the company. The
level bargaining and our measures of digitalization, i.e. 4% raise in magnitude of coefficients is slightly lower compared to OLS in Table 3, but still
significant and positive.
99
Table A4
Robustness check: inclusion of part-time for work organizational issues.
(1) (2) (3) (4)
b/se b/se b/se b/se
Share of college workers 0.176743*** 0.335029 0.175876*** 0.394366**

(0.037) (0.211) (0.037) (0.153)
Share of trained workers 0.044355*** 0.084889*** 0.043785*** 0.086722***
(0.013) (0.026) (0.013) (0.026)
Part time for work organization ¡0.011698 ¡0.037482 ¡0.011253 ¡0.039943
(0.015) (0.031) (0.015) (0.031)
Share of fixed-term workers ¡0.063452* 0.024923
(0.036) (0.099)
Second level agreement 0.037519** 0.116699** 0.036348** 0.117530***
(0.017) (0.047) (0.017) (0.041)
Firm size 0.069584*** 0.002455*** 0.070488*** 0.002269***
(0.005) (0.001) (0.005) (0.001)
VA per worker (log) 0.010045** 0.031856*** 0.009694** 0.031970***
(0.005) (0.010) (0.005) (0.010)
Share of workers over 50 − 0.041883 − 0.305846*** − 0.050584 − 0.299574***
(0.032) (0.068) (0.033) (0.068)
Share of workers 35–50 0.038186 − 0.038437 0.029231 − 0.036644
(0.028) (0.057) (0.029) (0.057)
Family firm − 0.000854 − 0.079207*** − 0.000001 − 0.083402***
(0.015) (0.029) (0.015) (0.028)
Share of workers with high school degree 0.097145*** 0.083543 0.095827*** 0.077018
(0.021) (0.055) (0.021) (0.047)
In a trade group 0.032799*** 0.074713*** 0.032598*** 0.076504***
(0.012) (0.026) (0.012) (0.025)
Management with tertiary education 0.030022* 0.107059*** 0.029831* 0.109131***
(0.018) (0.039) (0.018) (0.039)
Management with high school degree 0.009236 0.060184* 0.009560 0.060269*
(0.015) (0.036) (0.015) (0.035)
Female management 0.011439 − 0.015882 0.012198 − 0.014812
(0.017) (0.033) (0.017) (0.033)
Product innovation 0.066006*** 0.157441*** 0.065852*** 0.159106***
(0.013) (0.025) (0.013) (0.025)
Process innovation 0.081439*** 0.188607*** 0.081506*** 0.186201***
(0.013) (0.025) (0.013) (0.025)
Firm age 0.000143 0.000538** 0.000132 0.000532**
(0.000) (0.000) (0.000) (0.000)
FDI 0.089627*** 0.212873*** 0.089791*** 0.207190***
(0.028) (0.075) (0.028) (0.060)
Share of export on VA 0.000153 0.000575 0.000163 0.000521
(0.000) (0.000) (0.000) (0.000)
Observations 7746 7746 7746 7746
Non zero obs. 3719 3225
Wald Chi2 994.72 709.34 997.06 715.58
Prob > Chi2 0.0000 0.0000 0.0000 0.0000
Pseudo R2 0.1049 0.1052
100
a higher probability to introduce digital technologies as well as to follow classes. Consistently with the results presented Table 3, college-
an adoption pattern characterized by investment in multiple I4.0 tech graduates and trained workers have a positive and statistically signifi
nologies. Conversely, a higher share of older workers and family- cant effect on the adoption of I4.0 technologies for both SMEs and large
ownership are strongly and negatively correlated with the adoption of firms, and for firms operating in manufacturing and service sectors.
I4.0 technologies and also with ‘multi-technology’ strategies. Interestingly, the share of trained workers does not affect the probability
As for hypothesis 2, investment in training seem to favour technology of adoption for large firms, and the effect of second-level agreements is
adoption, validating the argument about the importance of tacit positive for SMEs but negative for larger firms. Decentralised bargaining
knowledge and the development of firm-specific capabilities that will seems to favour technology adoption in small and medium-sized orga
enable the extraction of productivity gains from the new investments. nisations, and not in larger firms typically more reliant on sector-level
Interestingly, however, the complementarity effect between skills and bargaining. This may indicate that in large businesses, second-level
digital technologies is stronger for skills generated through the educa bargaining might hamper the diffusion of I4.0 technology possibly
tion system than through on-the-job training. Results from Table 3 because of concerns among the labour force of increased control by
provide useful insights on the relation between firm-level work orga managers through technological surveillance. A more corporative atti
nization and the propensity to invest in new technologies. As discussed tude may prevail in small and medium-size firms, either fostering better
in Section 2, adoption goes hand in hand with organizational change. information exchanges about the technology within the company or
The use of ICTs has already shown patterns of complementarity with favouring a generally more cooperative context for shared technology
specific forms of work organisation involving decentralised decision- adoption decisions.
making and team working (Bresnahan et al., 2002). We find this effect All in all, these findings suggest that there is significant heteroge
also here, where our second hypothesis on the role of flexible staff ar neity in the population of adopters, but results are very similar between
rangements is confirmed.18 the manufacturing and service sectors, indicating that the new tech
Our results also indicate that the introduction of I4.0 technologies is nologies affect very different economic activities and have the potential
facilitated by the accumulation of knowledge through longer-term work to permeate not only manufacturing (the production model more closely
relationships rather than by the expectation of efficiency gains derived associated with the ‘smart factory’) but also services. Between the two
from the use of short-term contracts (in line with Kleinknecht, 2020). macro-areas, the variable that is significant for manufacturing but not
Results presented in column 1 of Table 3 show that firms with a higher for services is the role of second-level bargaining. This might signal
share of temporary jobs are on average less digitized than those char different patterns of governance for technology adoption decisions be
acterised by a higher share of permanent jobs. Finally, with respect to tween manufacturing and services, which may also be associated with
our final hypothesis, the role of second-level bargaining is also heterogeneous performance and employment effects post-adoption (see
confirmed and follows our expectations. The underlying mechanism is Cirillo et al., 2022b on recent qualitative evidence concerning automa
likely to be that more collective decision-making process, shared across tion technologies in services).
the firm’s hierarchical structure, favour riskier investment in new
technology. Note as well that conditioning on the probability of prior 5. Conclusions
investment, that is to say underplaying the role of first technology
adopters relative to persistent investors, the statistical significance of What firm characteristics favour the adoption of new production
flexibility and second-level bargaining disappears.19 technologies? This is a fundamental question if we are on the eve of a
In Table 4 we show the results of our econometric estimation on the Fourth Industrial Revolution and if the new wave of digital technologies
probability of adopting at least one I4.0 technology, by disaggregating is the backbone of the hyper-connected company of the future, charac
firms into size classes, i.e. firms with less than 250 employees vs. firms terized by strong interaction between new production technologies
with more than 250 employees, and the broad economic sector in which (smart production) and information and network infrastructures (smart
they operate, i.e. manufacturing and services.20 Given the attention paid services). While the average firm is still very far from this archetypal
to the role of firm size in the diffusion literature (see Section 2), it is model of production – and least in the Italian context – the process of
important to explore similarities and differences in behaviour across size diffusion of new enabling technologies is on its way and it is already
possible to identify some defining characteristics of adopters, even in a
context that is lagging behind in terms of digitalisation relative to
18
Table A4 displays results obtained using as work organizational variable the comparable economies.
main reason of applying part-time contracts. In detail, we create a dummy In this contribution we have exploited a large and unique dataset that
variable taking value of 1 if the company positively replies to the following includes fine-grained information about the technology adoption
option: ‘part-time better suits the company’s type of production and work choices made by firms. The vast majority of adopters opt for a single-
organisation’, and 0 otherwise. We did not find a statistically significant rela
technology, rather than an integrated (multiple-technology) approach.
tionship for work organizational part-time, although the share of fixed-term
The econometric evidence confirms a line of continuity with previous
employees remains significant and negative in the OLS equation.
19
As a robustness check, Table A5 in the Appendix shows the probabilities
studies of ICTs, with strong complementarities between skills and new
associated to the introduction of I4.0 technologies defined as internet of things, technology. Both human capital measured by education attainment
robotics, big data analytics, augmented reality and therefore excluding cyber levels and on-the-job training are positively associated with the adop
security. We still detect a significant and positive relationship between share of tion of digital technologies. Comparatively weaker evidence points to
college and trained workers and adoption of I4.0 technologies (not including the role of flexible work. Regarding this variable, some of the estima
cybersecurity). The implementation of Heckman correction weakens the sig tions indicate a negative, rather than positive, effect, pointing to the
nificance of estimates except for share of trained workers which remains importance of knowledge accumulation embodied in firm employees
strongly significant and positively associated to the adoption of I4.0. Overall rather than the efficiency gains of labour market flexibility. Decentral
cybersecurity is an intrinsic component of internet-based market transaction ised bargaining instead appears to favour new technology adoption,
and is susceptible to a broader pool of adopters that also include the large
albeit with strongly heterogeneous effects. The key implication for firm
number of service businesses – small and large – that compose the Italian
strategy is that workers’ skills are primary determinants of the adoption
economy across regions. More so than robots. This is corroborated by evidence
about the geographical spread of this technology, which is well-represented in of digital enabling technologies, arguably because they are necessary
all Italian macro-regions precisely because of its broader, not narrower, conditions for the extraction of productivity gains from the new assets.
‘enabling’ potential (Cirillo et al., 2022a). The corresponding policy lesson is that industrial policies that incenti
20
In the Appendix we provide a further exercise splitting the sample in vise digitization cannot only focus on the acquisition of assets but must
manufacturing and construction versus services (see Table A6). include strong components of upskilling and training, as well as
101
Table A5
Robustness check: alternative measure for I4.0 (excluding cybersecurity).
(1) (2) (3) (4)
At least one Number of I4.0 At least one Number of I4.0
b/se b/se b/se b/se
Share of college workers 0.071294** 0.089549* 0.031022 0.059362

(0.030) (0.049) (0.030) (0.051)
Share of trained workers 0.021218* 0.036154* 0.033721*** 0.008017
(0.011) (0.019) (0.011) (0.023)
Share of fixed-term workers ¡0.005948 ¡0.021650 0.012577 0.010702
(0.032) (0.055) (0.032) (0.056)
Second level agreement ¡0.000845 ¡0.016031 0.002944 0.013452
(0.013) (0.022) (0.014) (0.023)
Firm size 0.062436*** 0.001896*** 0.062393*** 0.000131***
(0.004) (0.000) (0.008) (0.000)
VA per worker (log) 0.013947*** 0.025616*** 0.014835*** 0.016289**
(0.004) (0.008) (0.004) (0.006)
Share of workers over 50 − 0.144781*** − 0.262581*** − 0.130679*** − 0.200965***
(0.029) (0.053) (0.040) (0.049)
Share of workers 35–50 − 0.041097* − 0.056013 − 0.037037 − 0.052279
(0.024) (0.043) (0.026) (0.043)
Family firm 0.000633 − 0.025490 − 0.014436 − 0.038484*
(0.012) (0.018) (0.011) (0.020)
Share of workers with high school degree 0.019363 0.018184 0.015643 0.008465
(0.018) (0.034) (0.018) (0.031)
In a trade group 0.016426 0.031029 0.027333*** 0.022164
(0.010) (0.019) (0.010) (0.020)
Management with tertiary education 0.011841 0.044168* 0.005996 0.012941
(0.015) (0.026) (0.014) (0.025)
Management with high school degree 0.007034 0.036393 0.008842 0.022084
(0.013) (0.024) (0.013) (0.022)
Female management 0.016933 − 0.001594 0.022128 0.021602
(0.014) (0.024) (0.014) (0.024)
Product innovation 0.052456*** 0.091830*** 0.066839*** 0.046937**
(0.011) (0.019) (0.010) (0.023)
Process innovation 0.077195*** 0.117249*** 0.074435*** 0.021422
(0.011) (0.019) (0.012) (0.030)
Firm age 0.000106 0.000248 0.000211* 0.000384*
(0.000) (0.000) (0.000) (0.000)
FDI 0.060700*** 0.102566*** 0.050175*** 0.058433**
(0.019) (0.035) (0.017) (0.030)
Share of export on VA 0.000190 0.000208 0.000217 0.000546*
(0.000) (0.000) (0.000) (0.000)
Sec. and reg. dummies Yes Yes Yes Yes
Observations 7746 7746 7675 7675
Non zero obs. 1755
Wald Chi2 1002.5 213.2 732.89 156.53
Prob > Chi2 0.0000 0.0000 0.0000 0.0000
Pseudo R2 0.1417
athrho 0.83 − 2.38
chi2(1) = 0.69 5.68
Prob > Chi2 0.4077 0.0171
appropriate policies for the supply of new skills through the institutional productivity and employment, and these should be investigated in a
formation of digital competences. variety of contexts of application. Further, it would be extremely
Far from being a policy evaluation design shedding light on the important to evaluate whether and how the implementation of new
effectiveness of the ‘National Enterprise Plan 4.0’, the evidence we models of production changes not only the internal but also the external
provide in this paper shows indirectly to what extent and under what organisation of the firm, and the related local vs. global development of
conditions the policy incentives hit the right targets. All in all, the evi value chains. Finally, one should investigate how and to which extent
dence shows that among the factors favouring digital technology macroeconomic conditions and heterogeneous demand can shape in
adoption, human capital and on-the-job training exert systematic posi vestment plans and mostly digitalization strategies. So far, the majority
tive effects. However, new microeconomic evidence on adoption be of Italian firms tends to adopt one technology, rather than many, it is
haviours is essential to shape strategy and policy decisions. This paper plausible that instances of multiple technology adoption would be much
contributes to filling an important gap in the literature on the emergent more complex to handle, and overall riskier and more costly. It is
Fourth Industrial Revolution, but of course many important questions therefore possible that firms opt for a gradual process of transformation,
remain unanswered. For instance, it will be fundamental to identify the quite independently from financial constraints hampering adoption
consequences of digitization on employment and wages, as the literature decisions. From this point of view, the techno-organizational capabil
has begun to do with a sharp, but perhaps too limited, focus on robots ities of firms and the type of strategic orientation versus technological
(Domini et al., 2021; Acemoglu et al., 2020). Given that the ongoing adoption both affect drivers and barriers to adoption (Cirillo et al.,
process of digitization includes a broad group of technologies and a 2022b). New qualitative, behavioural evidence could be very useful to
multiplicity of devices and techniques (as stressed by Balsmeier and gain a deeper and better understand of the economic mechanism at play
Woerter, 2019; Mina and Martinelli et al., 2021), it is possible that in the technological upgrading strategies of firms. This is even more
different technological combinations will generate different effects on crucial in the current phase of great uncertainty since the outbreak of the
102
Table A6 Robotics and enabling digital Technologies & Systems 4.0, Centro di
Robustness check: alternative grouping (manufacturing and construction). Competenza ai sensi dell’art. 1, c. 115, L. 11/12/2016, n. 232 e del DD di
At least one Concessione della DGPICPMI Prot. 09/2019 del 29/4/2019, CUP:
Manuf. + Construction B81J18000090008 and the support from the Italian Ministry of Educa
b/se tion, University and Research, PRIN-2017 project 201799ZJSN: “Tech
Share of college workers 0.251197*** nological change, industry evolution and employment dynamics.”
(0.075)
Share of trained workers 0.046604*** Appendix
(0.017)
Share of fixed-term workers ¡0.081110
(0.055) (Table A1, Table A2, Table A3, Table A4, Table A5, Table A6)
Second level agreement 0.051645**
(0.021) References
Firm size 0.079362***
(0.008) Acemoglu, D., 2002. Technical change, inequality and the labor market. J. Econ. Lit. 40,
VA per worker (log) 0.008277 7–72.
(0.006) Acemoglu, D., Restrepo, P., 2020. Robots and jobs: evidence from US labor markets.
Share of workers over 50 − 0.040482 J. Polit. Econ. 128 (6), 2188–2244.
(0.043) Acemoglu, D., Lelarge, C., Restrepo, P., 2020. Competing with robots: firm-level
Share of worker 35–50 0.031148 evidence from France. AEA Pap. Proc. 110, 383–388.
(0.039) Addison, J.T., Wagner, J., 1997. The impact of German works councils on profitability
Family firm 0.043538** and innovation. New evidence from micro data. Jahrbücher für Nationalökonomie
(0.020) und Statistik 216 (1), 1–20.
Share of workers with high school degree 0.074563*** Amemiya, T., 1985. Advanced Econometrics. Harvard University Press, Cambridge MA.
Antonietti, R., Antonioli, D., Pini, P., 2017. Flexible pay systems and labour productivity.
(0.027)
Int. J. Manpow. 38 (4), 548–566.
In a trade group 0.033532**
Appelbaum, E., Berg, P., 1999. High performance work systems: giving workers a stake.
(0.016)
eds.. In: Blair, M., Kochan, T. (Eds.), The New Relationship: Human Capital in the
Management with tertiary education 0.030316 American Corporation. Brookings Institution, Washington, DC, pp. 102–137.
(0.022) Archibugi, D., 2001. Pavitt’s taxonomy sixteen years on: a review article. Econ. Innovat.
Management with high school degree 0.020628 New Technol. 10, 415–425.
(0.018) Arvanitis, S., 2005. Computerization, workplace organization, skilled labour and firm
Female management − 0.007256 productivity: evidence for the Swiss business sector. Econ. Innovat. New Technol. 14
(0.024) (4), 225–249.
Product innovation 0.039440** Autor, D.H., Levy, F., Murnane, R. (2000). Upstairs, down- stairs: computer-skill
(0.017) complementarity and computer-labor substitution on two floors of a large bank.
Process innovation 0.109284*** NBER Working Paper 7890, Cambridge, Mass.
(0.017) Balsmeier, B., Woerter, M., 2019. Is this time different? How digitalization influences job
Firm age 0.000032 creation and destruction. Res. Policy 48 (8), 103765 article.
Bartelsman, E.J., Gautier, P.A., De Wind, J., 2016. Employment protection, technology
(0.000)
choice, and worker allocation. Int. Econ. Rev. (Philadelphia) 57 (3), 787–826.
FDI 0.069452**
Human capital revised Becker, G.S., Becker, G.S., 1994. Human Capital. A Theoretical
(0.033) and Empirical Analysis with Special Reference to Education. Chicago and
Share of export on VA − 0.000081 LondonThe Chicago University Press, pp. 15–28.
(0.000) Bloom, N., Sadun, R., Van Reenen, J., 2012. Americans do it better. Us multinationals
Sec. and reg. dummies Yes and the productivity miracle. Am. Econ. Rev. 102 (1), 167–201.
Observations 4513 Boothby, D., Dufour, A., Tang, J., 2010. Technology adoption, training and productivity
Wald Chi2 688.45 performance. Res. Policy 39, 650–661.
Prob > Chi2 0.0000 Böckerman, P., Johansson, E., Kauhanen, A., 2012. Innovative work practices and
Pseudo R2 0.1257 sickness absence. What does a nationally representative employee survey tell? Ind.
Corp. Change 21 (3), 587–613.
Bresnahan, T.F., Brynjolfsson, E., Hitt, L.M., 2002. Information technology, workplace
organization, and the demand for skilled labor. Firm-level evidence. Q. J. Econ. 117
Covid and the Russia-Ukraine conflict have likely reshaped the supply (1), 339–376.
chains of companies and their investment plans. Bresnahan, T.F., Trajtenberg, M., 1995. General purpose technologies ‘engines of
growth’? J. Econ. 65 (1), 83–108.
Bresnahan, T.F., 2010. General purpose technologies. eds. In: Bronwyn, H.H.,
CRediT authorship contribution statement
Rosenberg, N. (Eds.), Handbook of Economics of Innovation, Vol. 2. Elsevier,
pp. 761–791.
Valeria Cirillo: Conceptualization, Formal analysis, Writing – orig Brynjolfsson, E., McAfee, A., 2014. The Second Machine Age: Work, Progress, and
Prosperity in a Time of Brilliant Technologies. W.W. Norton & Company, New York.
inal draft. Lucrezia Fanti: Conceptualization, Writing – original draft.
Bugamelli, M., Cannari, L., Lotti, F., Magri, S., 2012. The Innovation Gap of Italy’s
Andrea Mina: Conceptualization, Writing – original draft. Andrea Production System. Roots and Possible Solutions, Questioni di Economia e Finanza n,
Ricci: Conceptualization, Data curation, Writing – original draft. 121. Banca d’Italia, Roma.
Calligaris, S., Del Gatto, M., Hassan, F., Ottaviano, G.I., Schivardi, F., 2016. Italy’s
Productivity Conundrum. A Study on Resource Misallocation in Italy. European
Data availability Commission, Directorate General Economic and Financial Affairs Publications Office
of the European Union, Luxembourg. Discussion Paper n.030.
The authors do not have permission to share data. Calvino, F., Criscuolo, C., Marcolin, L., Squicciarini, M., 2018. A taxonomy of digital
intensive sectors. OECD Science, Technology and Industry Working Papers, No. 2018/
14. OECD Publishing, Paris. https://doi.org/10.1787/f404736a-en.
Acknowledgments Calvino, F., et al., 2022. Closing the Italian digital gap: the role of skills, intangibles and
policies. OECD Science, Technology and Industry Policy Papers, No. 126. OECD
Publishing, Paris. https://doi.org/10.1787/e33c281e-en.
Valeria Cirillo acknowledges the support of the European Union’s Cameron, A.C., Trivedi, P.K., 2013. Regression Analysis of Count Data. Cambridge
Horizon Europe Research and Innovation programme under grant University Press.
agreement No 101069651 (BRIDGES 5.0). Views and opinions expressed Castro Silva, H., F, Lima, 2019. Technology, employment and skills. A look into job
duration. Res. Policy 46 (8), 1519–1530.
are however those of the author only and do not necessarily reflect those
Castellacci, F., 2008. Technological paradigms, regimes and trajectories: manufacturing
of the European Union. Neither the European Union nor HADEA can be and service industries in a new taxonomy of sectoral patterns of innovation. Res.
held responsible for them. Policy 37 (6–7), 978–994.
Andrea Mina acknowledges the support of ARTES 4.0 Advanced
103
Cetrulo, A., Cirillo, V., Guarascio, D., 2019. Weaker jobs, weaker innovation. Exploring Hall, B.H., Khan, B., 2003. Adoption of new technology. ed.. In: Jones, D.C. (Ed.), New
the effects of temporary employment on new products. Appl. Econ. 51 (59), Economy Handbook. Academic Press, San Diego.
6350–6375. Hamermesh, D.S., Pfann, G.A., 1996. Adjustment costs in factor demand. J. Econ. Lit. 34,
Cirillo, V., Fanti, L., Mina, A., Ricci, A., 2022a. Upgrading Italy’s industrial capacity: 1264–1292.
industry 4.0 across regions and sectors. SINAPPSI – J. Ital. Natl. Inst. Public Policy Heckman, J.J., 1979. Sample selection bias as a specification error. Econometrica 47 (1),
(INAPP) XI (2), 14–35. https://doi.10.53223/Sinappsi_2021-02-1. 153–161.
Cirillo, V., Rinaldini, M., Staccioli, J., Virgillito, M.E., 2021. Trade unions’ responses to Hoxha, S., Kleinknecht, A., 2020. When labour market rigidities are useful for
Industry 4.0 amid corporatism and resistance. Struct. Change Econ. Dyn. 56, innovation. Evidence from German IAB firm-level data. Res. Policy 49 (7), 104066.
166–183. Huselid, M.A., 1995. The impact of human resource management practices on turnover,
Cirillo, V., Rinaldini, M., Virgillito, M.E., Divella, M., Manicardi, C., Massimo, F.S., productivity, and corporate financial performance. Acad. Manag. J. 38 (3), 635–672.
Cetrulo, A., Costantini, E., Moro, A., Staccioli, J., 2022b. Case Studies of Automation ISTAT, 2017. Rapporto Sulla Competitività Dei Settori Produttivi. Edizione, Roma, 2017.
in Services. Publications Office of the European Union, Luxembourg, JRC129691. ISTAT, 2018. Rapporto Sulla Competitività Dei Settori Produttivi. Edizione, Roma, 2018.
https://doi.org/10.2760/347087, 2022, ISBN 978-92-76-58953-2. ISTAT, 2021. Rapporto Sulle Imprese2021. Struttura, Comportamenti e Performance Dal
Codogno, L., 2009. Two Italian puzzles. Are productivity Growth and Competitiveness Really Censimento Permanente. Edizione, Roma, 2021.
so Depressed? Working Paper no. 2. Ministero dell’Economia e Finanza, Roma. Jirjahn, U., Mueller, S., 2014. Non-union worker representation, foreign owners, and the
Cohen, W., Levinthal, D.A., 1990. Absorptive capacity. A new perspective on learning performance of establishment. Oxf. Econ. Pap. 66 (1), 140–163.
and innovation. Admin. Sci. Q. 35 (1), 128–152. Jovanovic, B., R.ousseau, P.L., 2005. General purpose technologies. eds. In: Aghion, P.,
Comin, D., Mestieri, M., 2013. If technology has arrived everywhere. Why Has Income Durlauf, S. (Eds.), Handbook of Economic Growth, Vol. 1, Part B. Elsevier,
Diverged?” NBER Working Paper, No. 19010. National Bureau of Economic Amsterdam, pp. 1181–1224.
Research, Cambridge, MA. https://doi.org/10.3386/w19010. Kagermann, H., Wahlster, W., Helbig, J., 2013. Securing the future of German
European Commission, 2018. Digital economy and society index 2018 Report. European manufacturing industry. Recommendations for Implementing the Strategic Initiative
Commission, Luxembourg. Industrie 4.0. Final report of the Industrie 4.0 Working Group. Acatech - National
Damiani, M., Pompei, F., Ricci, A., 2018. Family firms and labor productivity. The role of Academy of Science and Engineering, Munich.
enterprise-level bargaining in the Italian economy. J. Small Bus. Manag. 56 (4), Karshenas, M., Stoneman, P., 1993. Rank, stock, order and epidemic effects in the
573–600. diffusion of new process technologies: an empirical model. Rand J. Econ. 24,
Dauth, W., Findeisen, S., Südekum, J., Woessner, N. (2017), German robots - the impact 503–528.
of industrial robots on workers, centre for economic policy research (CEPR) Kleinknecht, A., 2020. The (negative) impact of supply-side labour market reforms on
Discussion Paper no. 12306, Available at SSRN: https://ssrn.com/abstract productivity. An overview of the evidence. Camb. J. Econ. 44 (2), 445–464.
=3039031. Kleinknecht, A., Van Shaik, F.N., Zhou, H., 2014. Is flexible labour market good for
David, P., 1969. A Contribution to the Theory of Diffusion. Stanford Center for Research innovation? Evidence from firm-level data. Camb. J. Econ. 38 (5), 1207–1219.
in Economic Growth, Stanford University, Stanford. Laursen, K., Meliciani, V., 2002. The relative importance of international vis-`a-vis
Davies, S., 1979. The Diffusion Process of Innovation. Cambridge University Press, national technological spillovers for market share dynamics. Ind. Corp. Change 11
CambridgeUK. (4), 875–894.
Devicienti, F., Manello, A., Vannoni, D., 2017. Technical efficiency, unions and Leiponen, A., 2005. Skills and innovation. Int. J. Ind. Org. 23 (n.5–6), 303–323.
decentralized labor contracts. Eur. J. Oper. Res. 260 (3), 1129–1141. Link, A., Siegel, D., 2003. Technological Change and Economic Performance. Routledge,
Domini, G., Grazzi, M., Moschella, D., Treibich, T., 2021. Threats and opportunities in London.
the digital era: automation spikes and employment dynamics. Res. Policy 50 (7), Lundvall, B.A., 2009. The future of innovation in the learning economy. eds.. In: Von
104137. Stamm, B., Trifilova, A. (Eds.), The Future of Innovation. Gower Publishing,
Dosi, G., 1988. Sources, procedures and microeconomics effects of innovation. J. Econ. Aldershot UK, pp. 40–41.
Lit. 26, 1120–1171. Malerba, F., 2002. Sectoral systems of innovation and production. Res. Policy 31 (2),
Dosi, G., 1991. The research on innovation diffusion. An assessment. eds.. In: 247–264.
Nakicenovic, N., Grübler, A. (Eds.), Diffusion of Technologies and Social Behavior. Mansfield, E., 1968. Industrial Research and Technological Innovation. W. W. Norton,
International Institute for Applied Systems Analysis (IIASA), Laxenburg, New York.
pp. 179–208. Marsili, O., Verspagen, B., 2002. Technology and the dynamics of industrial structure: an
Dosi, G., Marengo, L., 2015. The dynamics of organizational structures and performances empirical mapping of Dutch manufacturing. Ind. Corp. Change 11 (4), 791–815.
under diverging distributions of knowledge and different power structures. J. Inst. Martinelli, A., Mina, A., Moggi, M., 2021. The enabling technologies of industry 4.0:
Econ. 11 (3), 535–559. examining the seeds of the fourth industrial revolution. Ind. Corp. Change 30 (1),
Dosi, G., Marengo, L., Pasquali, C., 2006. How much should society fuel the greed of 161–188.
innovators? On the relations between appropriability, opportunities and rates of McGuirk, H., Lenhitan, H., Hart, M., 2015. Measuring the impact of innovative human
innovation. Res. Policy 35 (8), 1110–1121. capital on small firms’ propensity to innovate. Res. Policy 44 (4), 965–976.
Dosi, G., Nelson, R.R., Winter, S.G., 2000. The Nature and Dynamics of Organizational Menezes-Filho, N., Van Reenen, J., 2003. Unions and innovation. A survey of the theory
Capabilities. Oxford University Press, Oxford. and empirical evidence. eds.. In: Addison, J.T., Schnabel, C. (Eds.), International
Evangelista, R., 2000. Sectoral patterns of technological change in services. Econ. Handbook of Trade Unions. Edward Elgar Publishing, Cheltenham UK, pp. 293–334.
Innovat. New Technol. 9, 183–221. Metcalf, D., 2003. Unions and productivity, financial performance and investment:
Fabbri, R., Albano, Y., Curzi, T., 2018. Work autonomy, control and discretion in international evidence. eds.. In: Addison, J.T., Schnabel, C. (Eds.), International
Industry 4.0. eds.. In: Cantoni, F., Mangia, G. (Eds.), Human Resource Management Handbook of Trade Unions. Edward Elgar Publishing, Cheltenham UK, pp. 118–171.
and Digitalization. Routledge, London, pp. 111–130. Metcalfe, J.S., 1981. Impulse and diffusion in the process of technological change.
Fabiani, S., Schivardi, F., Trento, S., 2005. ICT adoption in Italian manufacturing: firm- Futures 13 (5), 347–359.
level evidence. Ind. Corp. Change 14 (2), 225–249. Metcalfe, J.S., 1998. Evolutionary Economics and Creative Destruction. Routledge,
Fairris, D., Askenazy, P., 2010. Works councils and firm productivity in France. J. Labor London.
Res. 31 (3), 209–229. Michie, J., Sheehan, M., 2003. Labour market deregulation, ‘flexibility’, and innovation.
FitzRoy, F.R., Kraft, K., 1990. Innovation, rent-sharing and the organization of labour in Camb. J. Econ. 27 (1), 123–143.
the federal republic of Germany. Small Bus. Econ. 2 (2), 95–103. Moro, A., Rinaldini, M., Staccioli, J., Virgillito, M.E., 2019. Control in the era of
Ford, M., 2015. The Rise of the Robots: Technology and the Threat of Mass surveillance capitalism. An empirical investigation of Italian Industry 4.0 factories.
Unemployment. Basic Books, New York. J. Ind. Bus. Econ. 46 (3), 347–360.
Freeman, C., Perez, C., 1988. Structural crises of adjustment. Business cycles and Mowery, D., Rosenberg, N., 1993. The influence of market demand upon innovation. A
investment behaviour. eds.. In: Dosi, G., Freeman, C., Nelson, R., Silverberg, G., critical review of some recent empirical studies. Res. Policy 22 (n.2), 107–108.
Soete, L. (Eds.), Technical Change and Economic Theory. Pinter Publisher, London, Nelson, R., Winter, S., 1982. An Evolutionary Theory of Economic Change. Harvard
pp. 871–901. University Press, Cambridge MA.
Freeman, R., Medoff, J., 1984. What Do Unions Do? Basic Books, New York. OECD, 2011. Skills for Innovation and Research. OECD Publishing, Paris.
Frick, B., Möller, I., 2003. Mandated works councils and firm performance. Labor Osterman, P., 1994. How common is workplace transformation and who adopts it? ILR
productivity and personnel turnover in German establishments. Schmollers Jahrbuch Rev. 47 (2), 173–188.
123 (3), 423–454. Ouadahi, J., 2008. A qualitative analysis of factors associated with user acceptance and
Garnero, A., Rycx, F., Terraz, I., 2019. Productivity and Wage Effects of Firm-Level rejection of a new workplace information system in the public sector: a conceptual
Collective Agreements. Evidence from Belgian linked panel data, Bonn. IZA model. Can. J. Admin. Sci. 25 (3), 201–2013.
Discussion Paper n.11568. Pavitt, K., 1984. Sectoral patterns of technical change: towards a taxonomy and a theory.
Genz, S., Bellmann, L., Matthes, B., 2019. Do German works councils counter or foster Res. Policy 13, 343–373.
the implementation of digital technologies? J. Natl. Stat. 239 (3), 523–564. Piva, M., Santarelli, E., Vivarelli, M., 2005. The skill bias effect of technological and
Giuri, P., Torrisi, S., Zinovyeva, N., 2008. ICT, skills, and organizational change: organisational change evidence and policy implications. Res. Policy 34, 141–157.
evidence from Italian manufacturing firms. Ind. Corp. Change 17 (1), 29–64. Prais, S.J., 1995. Productivity, Education and Training. Facts and Policies in
Gruber, H., Verboven, F., 2001. The diffusion of mobile telecommunications services in International Perspective. Cambridge University Press, Cambridge.
the European Union. Eur. Econ. Rev. 45 (3), 577–588. Raffaelli, R., Glynn, M.A., Tushman, M., 2019. Frame flexibility: the role of cognitive and
Guerrieri, P., Meliciani, V., 2005. Technology and international competitiveness: the emotional framing in innovation adoption by incumbent firms. Strat. Manag. J. 40,
interdependence between manufacturing and producer services. Struct. Change 1013–1039.
Econ. Dyn. 16, 489–502. Ramachandran, V., 1993. Technology transfer, firm ownership, and investment in
human capital. Rev. Econ. Stat. 75, 664–670.
104
Rogers, E.M., 1962. Diffusion of Innovations. Free Press of Glencoe, New York. Tether, B., Mina, A., Consoli, D., Gagliardi, D, 2005. A Literature Review on Skills and
Rubery, J., Grimshaw, D, 2003. The Organisation of Employment: An International Innovation. How Does Successful Innovation Impact on the Demand for Skills and
Perspective. Macmillan Education, UK. How do Skills Drive innovation. UK Department of Trade and Industry. Technical
Rosenberg, N., Rosenberg, N., 1976. Factors affecting the diffusion of technology. report.
Perspectives on Technology. Cambridge University Press, Cambridge, pp. 189–210. Toner, P., 2011. Workforce Skills and Innovation: an Overview of Major Themes in the
Schnabel, C., Wagner, J., 1994. Industrial relations and trade union effects on innovation Literature. OECD Directorate for Science, Technology and Industry (STI), Centre for
in Germany. Labour 8 (3), 489–504. Educational Research and Innovation (CERI). Technical report.
Seamans, R., Raj, M. (2018) AI, labor, productivity and the need for firm-level data, Winter, S.G., 1997. Knowledge and competences as strategic assets. Ed.. In: Klein, A.
NBER Working Paper 24239 http://www.nber.org/papers/w24239. (Ed.), The Strategic Management of Intellectual Capital. Elsevier Ed., New York,
Stoneman, P., 1981. Intra firm diffusion, Bayesian learning and profitability. Econ. J. 91, pp. 165–187.
375–389. Womack, J.P., Jones, D.T., Roos, D., 1990. The Machine that Changed the World: The
Stoneman, P., Battisti, G. (2010), The diffusion of new technology, in B.H. Hall and N. Story of Lean Production. Rawson Associates, New York, NY.
Rosenberg (eds) Handbook of the Economics of Innovation, Vol. 2, pp. 733–760, North Wooldridge, J.M., 2010. Econometric Analysis of Cross Section and Panel Data. MIT
Holland, Elsevier. Press, Cambridge: Massachusetts.
Teece, D.J., 2018. Profiting from innovation in the digital economy: enabling
technologies, standards, and licensing models in the wireless world. Res. Policy 47
(8), 1367–1387.
105

The Adoption of Digital Technologies - Investment, Skills, Work Organisation

Uploaded by

Copyright:

Available Formats

The Adoption of Digital Technologies - Investment, Skills, Work Organisation

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Adoption of Digital Technologies - Investment, Skills, Work Organisation

Uploaded by

Copyright:

Available Formats

Structural Change and Economic Dynamics 66 (2023) 89–105

Contents lists available at ScienceDirect

Structural Change and Economic Dynamics

The adoption of digital technologies: Investment, skills, work organisation

1. Introduction may or may not lead to a radically new ‘techno-economic paradigm’

Fig. 1. Percentage of firms investing in I4.0 technologies.

Fig. 2. Percentage of firms investing in I4.0 technologies.

Mechanical industry 57.2 12.5 14.7 7.0 3.9 46.9

Fig. 3. Education, training, temporary work by I4.0 investors.

Fig. 4. Share of firms signing second-level bargaining by number of I4.0 technologies.

Share of college workers 0.177028* 0.396379 0.115041* 0.266226

Observations 7746 7746 7675 7675

Share of college workers 0.177591* 0.382831 0.174657* 0.449092*

Observations 7746 7746 6774 6774

Share of college workers 0.176743* 0.335029 0.175876* 0.394366**

Share of college workers 0.071294** 0.089549* 0.031022 0.059362

You might also like