The unequal implications of Industry 4.0 adoption: evidence on productivity growth and convergence across Europe

Abstract

ABSTRACTDo new manufacturing technologies of the Industry 4.0 (I4.0) boost TFP growth? By adopting a distance-to-frontier framework, this paper explores whether the adoption of (advanced) digital technologies affect the sectoral TFP growth rates across manufacturing industries of 14 European countries, during the period 2009–2019. We rely on a novel measure of adoption of I4.0 technologies (namely, advanced industrial robots, additive manufacturing and industrial internet of things), exploiting highly detailed (8-digit level) information on imports of capital goods embodying such technologies. Our results suggest that adopting new digital manufacturing technologies of the I4.0 brings quantitatively important and statistically significant contributions to sectoral TFP growth rates, although these are mostly concentrated in countries close to the technology frontier. In turn, these technologies seem to have hampered the process of convergence between European technological leaders and laggards over the last decade.KEYWORDS: Industry 4.0fourth industrial revolutiontechnology diffusiontotal factor productivity (TFP)technological convergenceJEL CLASSIFICATION: O11O33O47 AcknowledgementsWe thank the Editor and two anonymous reviewers for their most constructive and helpful suggestions. The authors are especially grateful to Marco Grazzi and to Eduardo Ibarra-Olivo for their valuable comments and suggestions, and to the participants of the Italian Trade Study Group meeting (Ancona, November 2020) and of the seminar at the Economics Department (University of Perugia, May 2021) for their comments on earlier versions of this paper.Disclosure statementNo potential conflict of interest was reported by the author(s).Notes1 Their disruptive potential results from their potential for a widespread application across every manufacturing industry due to their ‘versatility and complementarity’ (Eurofound Citation2018, 3). Furthermore, while we already acknowledged the impact I4.0 technologies have on manufacturing operations – e.g. higher operational flexibility, higher production efficiency and quality, lower set-up costs and integration along the value chain, resulting in higher productivity and better performance overall (see also Skilton and Hovsepian Citation2017; Eurofound Citation2018) – additional high-level impact resides in the world of work and, in general, the entire society. On the one hand, a general concern around the ‘risks of new monopolies, mass redundancies, spying on workers, and the extension of precarious digital work’ (Davies Citation2015, 9) emerges. On the other hand, this transformation calls for a policy debate on the upcoming changes in the task content and occupational profiles of manufacturing employment (Frey and Osborne Citation2017; Eurofound Citation2018).2 In this work, we focus on studies addressing the implications of ‘physical’ I4.0 technology adoption. We stress the difference between ‘physical’ (i.e. capital embodied) and ‘digital’ (i.e. software-related) I4.0 technologies as such characteristic represents a crucial distinction, as also observed by Foster-McGregor, Nomaler, and Verspagen (Citation2019). In so doing, we intentionally avoid a detailed review of studies addressing the productivity implication of I4.0 technology development and innovation (e.g. patenting artificial intelligence; for a recent contribution, see Venturini Citation2022) as this would fall outside the purpose of our research. Notwithstanding, we redirect the reader to recent studies from Czarnitzki, Fernández, and Rammer (Citation2023) and Müller, Fay, and vom Brocke (Citation2018) who explore the productivity effects of ‘digital’ I4.0 technology adoption, i.e. artificial intelligence and big data, respectively.3 This measure is similar to the robot exposure index proposed by Acemoglu and Restrepo (Citation2020) to measure robot adoption at the local labour market level, used in several empirical studies, and to the import-weighted measures of I4.0 technology production proposed by Felice, Lamperti, and Piscitello (Citation2022) and Venturini (Citation2022).4 This information is computed by matching the 8-digit CN product codes for I4.0-related capital and intermediate goods with the corresponding 8-digit codes in Prodcom classification. In the Prodcom list, the first 4 digits of each product code coincide with the 4-digit NACE sector producing the good (Eurostat Citation2021).5 Data on I4.0 adoption measures (i.e. flows and stocks, aggregate and for each technology) are available upon request.6 We do not use sectoral PPPs, which would enable a more precise comparison across countries and sectors, since these are hardly available for all countries, sectors and years in our analysis. However, this is a lesser concern for our work as by using the within-groups estimator we should be able to filter out cross-country and cross-sector differences in prices.7 Country list: Austria (AUT), Belgium (BEL), Czech Republic (CZE), Germany (DEU), Denmark (DNK), Spain (ESP), Finland (FIN), France (FRA), United Kingdom (GBR), Italy (ITA), Netherland (NLD), Portugal (PRT), Slovak Republic (SVK), Sweden (SWE).8 Manufacturing industries list (NACE rev.2): 1 – Food products, beverages and tobacco (10–12); 2 – Textiles, wearing apparel, leather and related products (13–15); 3 – Wood and paper products; printing and reproduction of recorded media (16–18); 4 – Coke and refined petroleum products (19); 5 – Chemicals and chemical products (20); 6 – Basic pharmaceutical products and pharmaceutical preparations (21); 7 – Rubber and plastics products, and other non-metallic mineral products (22–23); 8 – Basic metals and fabricated metal products, except machinery and equipment (24–25); 9 – Computer, electronic and optical products (26); 10 – Electrical equipment (27); 11 – Machinery and equipment n.e.c. (28); 12 – Transport equipment (29–30); 13 – Other manufacturing; repair and installation of machinery and equipment (31–33).9 See, for instance, Cardona, Kretschmer, and Strobel (Citation2013) and Schweikl and Obermaier (Citation2020) for recent surveys of the literature on ICT and productivity.10 The model described in Section 3 assumes that it is not the identity of the technology frontier that is relevant in Equation (4), but the distance from the frontier itself, capturing the potential for technological catch-up. As the model allows for any country to switch endogenously from being a frontier to a non-frontier country and vice versa, only requiring that the lnDTF term correlates with the potential for technology transfer and productivity gains from catching-up. Thus, in column (8) we test an alternative specification of our model in which we measure lnDTF using the average TFP level for the two countries featuring the highest value as the frontier, and by computing ΔlnAF as the average growth rate between these two countries.11 While such result may be related to the yet mentioned lack of necessary conditions (e.g. a certain level of absorptive capacity) in the case of Slovakia and, to a certain extent, Portugal, the findings for Denmark may relate to the sectoral composition of the country, with a small and decreasing share of manufacturing as compared to services (similarly to other Nordic countries in our sample, i.e. Sweden and Finland).12 We followed Roodman (Citation2009) guidelines in the choice of the number of instruments.13 Computed following the growth accounting approach as described Stehrer et al. (Citation2019).14 According to estimates from Acemoglu and Restrepo (Citation2020), the average price of AIR ranges between 50,000 and 100,000 USD, while the average price for an industrial AM machine is between 200,000 and 250,000 USD according to our computations based on data from all major AM producers worldwide and reported by Senvol. Senvol’s data are available at http://senvol.com/machine-search/. Concerning IIoT, the total cost of deployment greatly varies depending on the sector and on the scale of the project. Using total cost of ownership (TCO) calculator for IoT applications by NOKIA, we estimate cost for a medium-sized factory to range between 1.6mln and 0.8mln USD. NOKIA’s IoT TCO calculator is available at https://pages.nokia.com/T007K9-Compare-Wireless-Critical-Connectivity-Options.