Teleworkability and the COVID-19 crisis: a new digital divide?

Sostero, Matteo; Milasi, Santo; Hurley, John; Fernández-Macías, Enrique; Bisello, Martina Working Paper Teleworkability and the COVID-19 crisis: a new digital divide? JRC Working Papers Series on Labour, Education and Technology, No. 2020/05 Provided in Cooperation with: Joint Research Centre (JRC), European Commission Suggested Citation: Sostero, Matteo; Milasi, Santo; Hurley, John; Fernández-Macías, Enrique; Bisello, Martina (2020) : Teleworkability and the COVID-19 crisis: a new digital divide?, JRC Working Papers Series on Labour, Education and Technology, No. 2020/05, European Commission, Joint Research Centre (JRC), Seville This Version is available at: https://hdl.handle.net/10419/231337 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ TECHNOLOGY LABOUR EDUCATION JRC Technical Report Teleworkability and the COVID-19 crisis: a new digital divide? JRC Working Papers Series on Labour, Education and Technology 2020/05 A Joint European Commission–Eurofound Report Matteo Sostero, Santo Milasi , John Hurley, Enrique Fernandez-Macías and Martina Bisello This Working Paper is part of a Working paper series on Labour, Education and Technology by the Joint Research Centre (JRC) The JRC is the European Commission’s science and knowledge service. It aims to provide evidence-based scientific support to the European policymaking process. The scientific output expressed does not imply a policy position of the European Commission. Neither the European Commission nor any person acting on behalf of the Commission is responsible for the use that might be made of this publication. Contact information Name: Fernandez-Macias Enrique Address: Joint Research Centre, European Commission (Seville, Spain) Email: [email protected] Tel.: +34 9544-87109 EU Science Hub https://ec.europa.eu/jrc JRC121193 Seville: European Commission, 2020 © European Union, 2020 The reuse policy of the European Commission is implemented by the Commission Decision 2011/833/EU of 12 December 2011 on the reuse of Commission documents (OJ L 330, 14.12.2011, p. 39). Except otherwise noted, the reuse of this document is authorised under the Creative Commons Attribution 4.0 International (CC BY 4.0) licence (https://creativecommons.org/licenses/by/4.0/). This means that reuse is allowed provided appropriate credit is given and any changes are indicated. For any use or reproduction of photos or other material that is not owned by the EU, permission must be sought directly from the copyright holders. All content © European Union 2020 How to cite this report: Sostero M., Milasi S., Hurley J., Fernández-Macías E., Bisello M., Teleworkability and the COVID-19 crisis: a new digital divide?, Seville: European Commission, 2020, JRC121193. Teleworkability and the COVID-19 crisis: a new digital divide? Matteo Sostero, Santo Milasi, John Hurley, Enrique Fernandez-Macias and Martina Bisello (European Commission JRC and Eurofound) Abstract The paper discusses the extent of teleworking in the EU before and during the COVID-19 outbreak, develops a conceptual analysis to identify the jobs that can be done from home and those that cannot, and on this basis quantifies the fraction of employees that are in teleworkable occupations across EU countries, sectors and socio-economic profiles. Using the occupational task descriptions provided in the Italian Indagine Campionaria delle Professioni, with additional indicators from the European Working Conditions survey, we estimate that 37% of dependent employment in the EU is currently teleworkable – very close to the estimates of teleworking indicated in real-time surveys during the COVID-19 crisis. Because of differences in the employment structure, the fraction of telewokable employment ranges between 33-44% in all but five EU member states. Even starker differences in teleworkability emerge between highand low-paid workers, between whiteand blue-collar workers, as well as by gender. Results suggests that that the large expansion of telework since the COVID-19 outbreak has been strongly skewed towards high-paid white-collar employment. Yet, enforced closures have likely resulted in many new teleworkers amongst low and mid-level clerical and administrative workers who previously had limited access to this working arrangement. This is consistent with the evidence showing that, beyond differences in the industrial and occupational structures, the pre-outbreak large differences in telework prevalence across EU countries were largely driven by other factors, notably the organisation of work, regulation, and management culture. This paper also discusses some policy implications that the current experience of telework may have for the future of work. Keywords: Teleworking; remote work; work from home; tasks; COVID-19 Teleworkability and the COVID-19 crisis: a new digital divide? Authors: Matteo Sostero, Santo Milasi, John Hurley, Enrique Fernandez-Macias and Martina Bisello (European Commission JRC and Eurofound) Acknowledgements: This paper was produced within the COVID & Empl Working Group, composed by researchers from the JRC, Eurofound, Cedefop and EU-OSHA, including Martina Bisello, Maurizio Curtarelli, Marta Fana, Enrique Fernández-Macías, John Hurley, Santo Milasi, Joanna Napierala, Annarosa Pesole, Konstantinos Pouliakas, Ignacio González-Vázquez, Matteo Sostero, Songül Tolan, Sergio Torrejón, Cesira Urzi Brancati, Simon Walo. The authors would like to thank Maurizio Curtarelli, François Dessart, Ignacio González Vázquez, Irene Mandl, Dimitris Mavridis, Dimitrios Pikios and ESCO team, Konstantinos Pouliakas, Sergio Torrejón Pérez, Anneleen Vandeplas, René Van Bavel, Oscar Vargas Llave for useful comments to earlier versions of this paper. Joint Research Centre reference number: JRC 121193 Teleworkability and the COVID-19 crisis: a new digital divide? Contents Abstract ............................................................................................................................................................................................................... 1 Contents ............................................................................................................................................................................................................. 3 Introduction ....................................................................................................................................................................................................... 5 Working from home in the EU before the COVID-19 pandemic: what we know ......................................................... 7 How common was telework in the EU before the COVID-19 pandemic? And for whom? ............................... 8 Where was telework more widespread before the COVID-19 crisis? And why? ................................................. 14 Telework and working conditions before COVID-19 pandemic ................................................................................... 19 Work from home during the COVID-19 pandemic: first evidence from real-time surveys ........................... 21 What jobs can be done from home – a tasks approach ....................................................................................................... 28 Teleworkability as a technical feasibility: a conceptual framework for occupational analysis .................. 28 Constructing indices of teleworkability ..................................................................................................................................... 31 Technical teleworkability index ................................................................................................................................................. 31 Social interaction index ................................................................................................................................................................. 33 Additional indices ............................................................................................................................................................................. 34 Advantages and limitations of the methodology ................................................................................................................ 34 What jobs are teleworkable? ................................................................................................................................................................ 38 Teleworkability indices of occupations ..................................................................................................................................... 38 Teleworkability across the EU ........................................................................................................................................................ 45 Discussion ....................................................................................................................................................................................................... 52 Teleworkability and work organisation ..................................................................................................................................... 52 Some considerations on telework after the lockdown ..................................................................................................... 53 Conclusions .................................................................................................................................................................................................... 58 References...................................................................................................................................................................................................... 60 Annexes ............................................................................................................................................................................................................ 63 Constructing the technical teleworkability index ................................................................................................................ 63 Values of technical teleworkability and social interaction indices............................................................................ 66 Teleworkability and the COVID-19 crisis: a new digital divide? Teleworkability and the COVID-19 crisis: a new digital divide? 5 Introduction The telework revolution has been predicted intermittently for over a generation - without ever arriving. Already in 1980, Alvin Toffler’s The Third Wave popularised the idea of the ‘electronic cottage’ which would ’shift literally millions of jobs out of the factories and offices into . . . where they came from originally: the home’ (Toffler, 1980, cited in Messenger, 2019). The terms ‘telework’ and ‘tele-commuting’ were reportedly coined in 1976 by Jack Nilles in “The TelecommunicationsTransportation Tradeoff,” where he and co-authors argued that the congestion problem plaguing American cities could be solved by de-centralising office activities into smaller hubs, and communicating between these hubs through human messengers and mainframe computers (Newport, 2020). At the time of Nilles’ writing, personal computers were not widespread; they came of age in the 1980s and their connection to the internet from the mid-1990s would lay the necessary conditions for a broader take-up of telework. For many workers, especially those in in knowledge-based sectors, these new technologies would free employees to carry on their work remotely, unshackled from their workplaces at least some of the time. The benefits would include more leisure time, enhanced work-life balance, greater work autonomy at the individual level, reduced congestion and environmental pollution at the societal level. But the reality has largely been disappointing so far. The most recent estimates from representative household surveys indicate that, four decades later, teleworking accounted for a relatively marginal share of paid labour in the EU, with fewer than one in twenty employees reporting working in this way regularly in 2018, and less than one in ten occasionally. This all changed, abruptly and of necessity, in the first semester of 2020 as a result of public health measures designed to stem the spread of COVID-19. In the extraordinary circumstances of a viral pandemic, with many workplaces in enforced closure, a vast ad-hoc social experiment took place in which teleworking became the customary mode of working for many employees with hitherto limited or no experience of working in this way. This is notably the case for those carrying out office-based work, especially in service sectors, who are largely reliant on intensive networked computer use. In the main, these are sectors where workers enjoy many labour market advantages – less physically arduous working conditions, higher pay, greater job security (Adams-Prassl et al, 2020). These advantages have been sharpened by COVID-19 as such work – and the businesses and employment relationships that depend on such work – has proved more resilient than much customer-facing service work, in sectors like air travel, tourism and hospitality, and restaurants. In this paper, we quantify the extent of teleworking before and during the COVID-19 outbreak, develop a broader conceptual analysis of teleworkability and sketch how the present experience may precipitate lasting changes in work organisation, who the beneficiaries of such changes may be and what issues such changes may raise for workers, employers and policy-makers. The paper explores these issues focusing mostly on the preand post-outbreak trends in telework among dependent employees. That is because dependent employees, more than the self-employed, are those who have experienced the largest increase in teleworking because of the COVID-19 outbreak, while also being more likely to witness long-lasting changes to their work organisation looking forward. Indeed, the need to work from home induced by the outbreak has removed, at least temporarily, many of the barriers that may have limited the adoption of telework over the past decade – e.g., regulation, norms of office work, and the unwillingness of employers and managers to extend unsupervised autonomy, driven by a lack of trust. To what extent these barriers may decline further in the future remains uncertain. The paper is structured as follows. In a first section, we present data on teleworking incidence preCOVID-19 from representative European surveys as well as first indications of its expanded incidence post-COVID-19 based on real-time survey data. First estimates show that the share of those teleworking regularly has increased anywhere from 3-5% to a third or more of employees at Teleworkability and the COVID-19 crisis: a new digital divide? 6 EU level, with significant variations across member states, and between North and South, and between East and West. This huge expansion in telework calls for a reflection on what exactly it is that makes some occupations amenable to it, while others are not, beyond the specific contingencies of pandemicinduced social reorganisation. In a second section, we develop and operationalise our hypothesis. Based on a conceptual framework of work tasks and on the state of current technology, we argue that the ultimate determinant of occupational teleworkability is the lack of physical handling tasks. Occupations rich in such tasks – those for example of nurses, of production line workers in manufacturing, of farmers – simply cannot be performed remotely with the available technologies. In principle, all other jobs may be performed remotely or from home, with varying degrees of difficulty. There are additional factors that either constrain or qualify the ability to telework – such as the extent of social interaction required in a job – while other factors facilitate it – such as the infrastructure for ICT connectivity in computer-facing knowledge-based work. These are discussed in the text, but it is our contention that the main impediment to teleworkability are physical handling tasks that mean a job can only reasonably be performed in specific locations. This follows in part the insight of the first analysis of teleworkability published soon after the onset of the crisis (Dingel and Neiman, 2020) and tries to provide a firmer conceptual underpinning for that insight based on an existing job tasks framework (Fernandez-Macias and Bisello, 2020). The approach taken is similar to that of earlier research designed to measure the extent of either “offshorability” (Blinder, 2009) or of “automatability” (Frey and Osborne, 2017) of jobs in advanced economies, where the task characteristics of jobs are used to proxy their exposure to potentially disruptive forms of change. Both widely-cited works were criticised regarding the high estimates of vulnerable employment and imprecise time-scale of the effects. The rejoinder of the authors was in both cases that their estimates were based on a technical potential – of offshorability, automatability – rather than predictive forecasts of actual offshoring or actual automation of jobs. This was prudent, as the incidence of actual offshoring or mass automation and consequent employment displacement in advanced economies largely has failed to materialise. By contrast, the change considered in this paper – in this case, mass telework – was not expected to occur so suddenly (Schwind and Llave, 2019). During the COVID-19 lockdown, all or nearly all jobs that were teleworkable in late April 2020 were being performed from worker’s homes. In a third section, we present a classification of occupations in terms of technical teleworkability. This is based on a job-tasks framework and assigns teleworkability values to over 130 occupations (ISCO 3 digit) based on the conceptual approach previously described. The main source for this classification is the detailed occupational task descriptions from the Italian Indagine Campionaria delle Professioni with additional indicators from the European Working Conditions survey. Based on this classification, we then assess the shares of teleworkable employment based on the classification developed. We estimate that around 36% of dependent employment in the EU is currently teleworkable, very close to the estimates of teleworking indicated in real-time surveys during the COVID-19 crisis. There are variations in incidence by member state based on differences in employment structures, even starker differences between well-paid workers and low-paid workers, between white-collar and blue-collar workers but also – perhaps more surprisingly – by gender. A fourth section discusses how teleworking might develop following the current expansion. It highlights some policy implications in terms of work organisation, regional and international inequalities, an occupational health and safety. A final section summarises and concludes. Teleworkability and the COVID-19 crisis: a new digital divide? 13 Box 3: Disparities in access to telework reflect other dimensions of labour market inequality. Data from the EU survey on ICT usage among workers allow to complement our understanding of the socio-economic profile of the typical teleworker before the onset of the COVID-19 crisis.3 These data show that close to 40% of workers who completed tertiary education worked from home with some frequency, against an average of 10% of those with secondary education and around 3% for those with low or no education (see Figure 6 below). Similarly, around 25% of workers in the top quartile of the EU-27 income distribution have access to telework – a share that declines to less than 10% among those in the bottom half. Telework was also slightly more common among those in standard employment contracts, with higher shares for those with permanent and full-time work contracts. Interestingly, workers with children teleworked slightly more than people without them. This may suggest that employers and employees have, at least in part, been using telework to balance work and family life. Finally, echoing findings stemming from the EU LFS data, no significant gender differences in the prevalence of telework are detected in the EU survey on ICT usage among workers Figure 6: The socio-economic profile of teleworkers Note: Figures refer to the share of employee working from home by frequency of working from home. Source: Eurostat, ICT usage among workers survey (variable code: isoc_iw_hem) 3 Figures on the overall prevalence of working from home from the EU survey on ICT usage among workers are slightly larger than those retrieved from the EU LFS, and the most recent European Working Conditions survey (2015). In fact, according to this survey, 5% of EU-27 employees reported working from home every day or almost every day in 2018, 7% reported doing so at least once a week (but not every day), while another 10% reported working from less than once a week. Teleworkability and the COVID-19 crisis: a new digital divide? 14 Where was telework more widespread before the COVID-19 crisis? And why? The ability to effectively adapt to large-scale teleworking is likely to differ significantly across EU countries. Pre-outbreak, there were large, and growing, differences in the prevalence of telework across its Member States. As of 2019, the share of employees working from home regularly or at least sometimes was above 25% in most Northern European countries, including Sweden, Finland, and the Netherlands, whereas it was below 10% in 15 of the 27 EU Member States (see Figure 7, below). Between these two extremes, there were countries such as Belgium, France and Portugal where the share of telework ranged between from 13 and 19%. Countries in Northern Europe are also those with the largest growth in the prevalence of telework since 2009, albeit sizable increases also took place in a number of other Member States including Portugal, Estonia, and Ireland. Figure 7: Prevalence of telework across EU Member States in 2009 and 2019, (% of employees) Note: Data on the frequency of telework in 2009 are not available for the Netherlands and Cyprus. Source: Eurostat, LFS. (variable code: lfsa_ehomp). Differences in industrial structures is one of the main factors explaining varying prevalence of telework across EU countries. Countries such as Sweden, Finland, and the Netherlands, where workers in knowledgeand ICT-intensive service sectors account for a larger share of total employment are not surprisingly those where telework is more widespread (see Figure 8, next page). Teleworkability and the COVID-19 crisis: a new digital divide? 15 Figure 8: Industrial structure of employment and telework Note: Knowledge and ICT intensive sector include: Publishing activities, Motion picture, video and television programme production, sound recording and music publishing activities Programming and broadcasting activities, Telecommunications, Computer programming, consultancy and related activities; Information service activities, Legal and accounting activities, Activities of head offices; management consultancy activities, Architectural and engineering activities; technical testing and analysis, Scientific research and development, Advertising and market research, Other professional, scientific and technical activities, Education, Human health activities, Creative, arts and entertainment activities. Source: Authors ’calculations based on Eurostat, LFS. Yet, even across knowledgeand ICT-intensive services, there are significant difference across countries. Figure 9 shows that, before the outbreak of the pandemic, differences in the share of teleworkers across EU countries were sizable even within the same sector. For instance, while more than 50% of workers in knowledge-intensive business services were, usually or sometimes, teleworking in Sweden and the Netherlands, this share was below 25% – the EU-27 average – in Austria (22%), Germany (17%), and Italy (4%). Similar degrees of cross-country heterogeneity in the prevalence of telework can be observed in education, IT and communication, and to a lesser extent in administrative and support services. Overall, this evidence suggests that, beyond differences in the industrial structure, the heterogeneous prevalence of telework across EU countries may significantly depend on other within-sector differences, including differences in the occupational mixes, the distribution of employment by firm size, the rate of self-employment, workers’ and firms’ affinities with digital technologies, as well as organisation and management cultures. Different regulatory frameworks (legislation, collective agreements) may also support or hinder deployment (Vargas Llave and Weber, 2020). Looking forward, all these factors are likely to shape the pace, scale, and effectiveness of a wider adoption of teleworking across sectors, firms, and in turn countries. Teleworkability and the COVID-19 crisis: a new digital divide? 16 Figure 9: Prevalence of telework by country in selected sectors, 2018 Note: see note to Figure 4 for a detailed definition of the sectoral groups. Source: Authors’ calculations from ad-hoc extractions of EU-LFS data provided by Eurostat. Box 4: Differences across countries in the prevalence of telework: Decomposing between and within variation of industries and occupations. Before the outbreak, there were large differences in the prevalence of telework across countries, as shown in Figure 6. These cross-country differences, that can be measured using the EU-LFS as our source, are mostly explained by: i) Differences in the industrial and occupational structures of employment, ii) Differences in characteristics within sectors and occupations, such as the occupational composition of the workforce in a given sector, and the way the work of each occupation is organised. In order to provide a first assessment of the relative importance of these factors in explaining the differences in the prevalence of telework across countries, we compare the share of telework across EU countries to that of Sweden, the reference country, and the related difference is decomposed into differences in the industrial/occupational structure of employment (between component) and differences within industries/occupations (within component) (for a similar decomposition approach see Nedelkoska and Quintini, 2018). We decompose the change in the prevalence of telework according to the following formula: ∆𝑇𝑇𝑐𝑐=��∆𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖,𝑐𝑐≠𝑆𝑆𝑆𝑆 ∗ 𝑇𝑇𝑖𝑖,𝑆𝑆𝑆𝑆� 𝑖𝑖 + ��𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖,𝑆𝑆𝑆𝑆 ∗∆𝑇𝑇𝑖𝑖,𝑐𝑐≠𝑆𝑆𝑆𝑆� 𝑖𝑖 The total difference in the prevalence of telework (∆𝑇𝑇𝑐𝑐) between any country (𝑐𝑐) different from Sweden (𝑐𝑐 ≠ 𝑆𝑆𝐸𝐸) and Sweden can be decomposed into a between sector/occupation component ∑(∆𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖,𝑐𝑐≠𝑆𝑆𝑆𝑆 ∗ 𝑇𝑇𝑖𝑖,𝑆𝑆𝑆𝑆) 𝑖𝑖 and a within sector/occupation component ∑�𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖,𝑆𝑆𝑆𝑆 ∗∆𝑇𝑇𝑖𝑖,𝑐𝑐≠𝑆𝑆𝑆𝑆� 𝑖𝑖. Teleworkability and the COVID-19 crisis: a new digital divide? 17 The term 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖 is the employment share in a specific industry and occupation-specific, and 𝑇𝑇𝑖𝑖 is the prevalence of telework in a specific industry and occupation; 𝑖𝑖 indicates the sector/occupation. Figure 10 and Figure 11 below show the results of this decomposition for industries and occupations respectively. Differences in the industrial and occupational structures do matter in explaining the overall differences in the spread of telework across countries – as indicated by the contribution of the between component for sectors and occupations shown in Figures 10 and 11 below, respectively. However, for the majority of the EU countries, the within component explains most of the deviation in telework rates from Sweden, both along the sectoral and the occupational dimension. For instance, the within-sectors variance accounts for 60% of more of the total difference in countries such as Greece, Poland, Hungary and Italy, explaining even larger fractions in Germany or Austria. A broadly similar picture emerges when the prevalence of telework is decomposed along the occupational dimension. Taken together, these findings suggest that the differences in the uptake of telework between Sweden and most of the other EU countries primarily depend on different rates of telework in the same sector and occupation, and only secondarily on the fact that Sweden has a larger share of employment in telework-compatible sectors and occupations. This largely reflects the fact that, with respect to Sweden, many EU countries show smaller shares of telework-compatible occupations within any given sector, while also organising and managing the work of each occupation in a way that is less conducive to telework. Conversely, when Sweden is compared to other Northern European countries, such as Finland and Denmark, the contribution of within component remains rather modest. That is because Sweden and the other Northern European countries tend to have quite similar occupational structures within any given sector and to adopt similar forms of work organization. As a result, differences in rates of telework among these countries are largely driven by differences in the sectoral and occupational structures. Figure 10: Sectoral decomposition of the differences across countries in the prevalence of telework Source: authors’ calculations from ad-hoc extractions of EU-LFS data provided by Eurostat Teleworkability and the COVID-19 crisis: a new digital divide? 18 Figure 11: Occupational decomposition of the differences across countries in the prevalence of telework Source: authors’ calculations from ad-hoc extractions of EU-LFS data provided by Eurostat Larger companies are more likely to adopt flexible work arrangements (including teleworking) than smaller ones. For instance, the share of teleworkers in knowledge-intensive business services is higher in countries such as the Netherlands, Sweden, and Finland where between 30 and 40% of the workforce in the sector is employed in firms with 50 employees or more (see Figure 12, next page). Conversely, this share is considerably lower in countries such as Italy and Croatia, where firms with 50+ employees account for less than 15% of employment in knowledge-intensive business services. There are of course some notable exceptions to this pattern. For example, the share of teleworkers in these sectors is relatively high in Greece and Slovenia despite the very low concentration of employment in mediumand large-sized enterprises, whereas the opposite is true in Germany, and to a lesser extent in France. Teleworkability and the COVID-19 crisis: a new digital divide? 19 Figure 12: Telework and employment by firm size, knowledge-intensive business services, 2018 Note: see note to Figure 4 for the definition of knowledge-intensive business services Source: Authors’ calculations from Eurostat, LFS. Telework and working conditions before COVID-19 pandemic The previous subsection showed where home-based telework was most common in the EU before the COVID-19 crisis, across countries, sectors, and occupations. Using data from the European Working Conditions Survey (EWCS), this section sheds light on the pre-outbreak working conditions of teleworkers – following our definition, those who reported working from home at least some of the time – and how these differ from those of employees who never worked from home. The main advantage of the EWCS over the EU-Labour Force Survey (LFS) is its broader coverage of dimensions on quality of work, such as working time, worker autonomy, job satisfaction and health outcomes. It also includes useful details on the methods and tools of work – including computer use – which may be correlated with the ability to work remotely. This information helps us describe the conditions of those working from home, beyond the demographic and job-related variables already indicated. Before the COVID-19 crisis, those who regularly worked from home full-time put in longer hours – 40.8 hours per week on average, compared to 39.8 hours per week for those who never worked from home. Even greater differences have been identified in other studies of European teleworkers (Schwind and Vargas, 2019), who found that the share of teleworking and mobile working employees working very long hours (>48 hrs per week) is much greater than that of nonteleworking counterparts, especially among male employees. Teleworkers also spent more time commuting to work (46-50 minutes per day compared to 40 minutes for those who never work from home). Perhaps those with longer commutes were also more likely to apply for homeworking. There is no significant difference between those who worked from home at least occasionally and those who did not, in terms of subjective indicators like job satisfaction, or their satisfaction with work-life balance (measured by the question ‘how do your Teleworkability and the COVID-19 crisis: a new digital divide? 20 working hours fit in with your family or social commitments outside work?’). There is also no difference in terms of proxy indicators for work intensity, such as working to tight deadlines or at high speed. Those who worked from home at least occasionally are more likely to report that their work affects their health positively, though they are also somewhat more likely to report higher levels of stress (14% of them report feeling stress at work ‘all of the time’, compared with 9% of those who never work from home). One strong difference is in the lack of boundary between work and home life – 37% of regular homeworkers, 18% of occasional homeworkers but only 4% of those who homework ‘less often’ or ‘never’ report having to work in their free time either daily or several times a week. In relation to work content, teleworkers were much more likely to report that their work involves high levels of autonomy and is cognitively rich (involves learning new things or complex tasks), which is unsurprising given their occupational and skills profile as discussed in the previous section. The survey also shows that computer use is correlated with the telework incidence, though the correlation is lower than might be expected. Those who worked from home (notably on an occasional basis) are highly likely to report intensive computer use in their work (working with computers ‘all of the time’ or ‘nearly all of the time’). Yet, intensive computer use is also reported by employees who never worked from home. This suggests that, among all the factors driving telework, computer use is an enabling technology, but ultimately matters less than the task content of one’s occupation, or the position in the organisation. Perhaps a more surprising finding is the extent of regular social interaction, including activities like dealing with non-colleagues such as clients, pupils, and patients, among those who work from home at least sometimes. One would expect that social interaction is an impediment to working from home, under the assumption that the social tasks involved are in many cases more difficult (or even impossible) to be carried out from home. But this appears not to be the case in practice. It may be that some of those social tasks are apt to be performed remotely using existing technology – e.g. via phone, email, or video-conferencing – without necessarily losing quality. An alternative explanation could be that teleworkers are able to compartmentalise tasks that require face-to-face interaction with non-colleagues in those periods when they are not teleworking. This ability to compartmentalise one’s work into teleworkable and non-teleworkable tasks is itself evidence of a high degree of work autonomy. As we have seen, a large share of occasional telework is carried out by managers and professionals, occupations both rich in social interaction and with high levels of work autonomy. The European Working Conditions Survey paints a picture consistent with the EU Labour Force Survey of the cross-country variation in the extent of working from home arrangements, with a high incidence in the Benelux and Nordic member states and a relatively lower one in Germany, Italy and Spain. It also presents a similar profile of the archetypal teleworker – more experienced, autonomous, highly educated, well-paid employees, working in higher level occupations, often as knowledge workers in services. Such a profile applies in particular to those who occasionally work for home, where this condition appears to be partly a privilege associated to high professional status, in part a compensation for longer hours working and blurred boundaries between professional and personal life. Teleworkability and the COVID-19 crisis: a new digital divide? 21 Work from home during the COVID-19 pandemic: first evidence from real-time surveys In response to the COVID-19 pandemic, several surveys have tried to estimate how many people were working from home, following the confinement measures that restricted mobility and led to the shuttering of many workplaces. In this section, we summarise results from some ad-hoc, realtime surveys. These sources can fill the gap resulting from an obvious weakness of official household or periodical surveys in covering unanticipated and fast-moving events such as a pandemic and its effects – the time lag between data collection and publication4. EU-LFS quarterly data is only published 3-4 months after the end of the quarter in question. It will be mid-summer 2020 at the earliest before the beginnings of the employment impacts of the crisis are evident in the EU-LFS. Each of the surveys cited below 5 has been mobilised quickly, in real time and online, either using an existing panel of respondents and with a claim to representativeness in the countries covered (University of Mannheim, University of Oxford, US Federal Reserve Board, UK ONS), or using an open online survey (Eurofound, Whitaker Institute/NUIG). While post-stratification weights were used in the latter surveys to adjust samples to known population shares for standard demographic variables, biases in the raw samples (disproportionately large female and tertiaryeducated shares of respondents, for example) suggest that some caution should be taken in interpreting the results. The most extensive real-time survey source in Europe is Eurofound’s COVID-19 survey, an online survey initiated in April 2020 which received nearly 62,000 completed responses in the EU member states in a first round (fieldwork: April 7th-April 30th)6. Of interest to this paper, in a small battery of work-related questions it asked on whether respondents had taken up teleworking / working from home following COVID-19 confinement measures and whether they had previous experience doing so. In the Eurofound study, over a third (39%) of employees currently working in the EU started to work from home following the pandemic, compared to 20% who indicated working from home at least ‘several times a month’ pre-COVID. Those who were working from home during the pandemic are mainly employees who had regular previous telework experience (50%) compared to 25% of those who had no previous telework experience. The share of those currently working from home was over 30% in all but four Member States (see Figure 13, next page). Nonetheless, the range of incidence between countries was wide (from 18% in Romania to 59% in Finland). The highest proportions of employees working from home were in the Nordic and Benelux countries, reflecting findings from the EWCS and LFS on the relatively high pre-outbreak prevalence of telework in these countries. 4 The EWCS is a five yearly survey. By unlucky timing, the COVID-19 crisis has interrupted data collection of the seventh wave (2020) after around a quarter of the survey responses had been completed. 5 The survey findings summarised here are principally from the following sources: the Mannheim Corona survey (for Germany; see Mohring et al, 2020), the COVID inequality survey (covering UK, USA and Germany; see Adams-Prassl et al, 2020) and Eurofound’s online COVID-19 survey (covering EU27 + Norway + UK; see Eurofound, 2020). Some non-EU findings are also presented for broader context; US findings from a Federal Research Board supplementary survey carried out in April 2020 (FRB, 2020) and UK findings from a similar exercise in the same month (ONS, 2020). 6 First findings were published on 7th May 2020. The survey was an open online survey, respondent recruitment was via snowballing and social media advertisements, the target population were those aged 18 and over in the EU member states and the UK. Data was weighted by age crossed with gender (in 12 age-gender combinations), urbanisation level (2 categories) and education level (2 categories) to reflect demographic composition in each country (Ahrendt et al, 2020; Eurofound, 2020). The survey asked respondents a variety of questions on living and working conditions during the COVIDconfinement. Respondents were not asked to indicate occupation (ISCO) or sector (NACE) in which they worked but a follow up wave 2 of the survey planned for early summer 2020 will include such questions and will also include a specific module of eight questions on teleworking. Teleworkability and the COVID-19 crisis: a new digital divide? 22 Women were somewhat more likely to report having started working from home post-outbreak than men (41% vs 37%, see Figure 14 below), whereas the opposite was true pre-outbreak, when 21% of men worked from home at least several times per month against 18% of women. The biggest rise in the prevalence of working from home was amongst younger employees supporting a narrative that the COVID-19 crisis has equalised access and neutralised at least provisionally the status-related dimension of access to telework. Figure 13: Employees working from home during COVID-crisis, by country % Source: EF COVID survey. Note: no data for Slovenia. Figure 14: Working from home during COVID-19 crisis, EU27%: work and personal characteristics Source: EF COVID survey. Note: no data for Slovenia. 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70 15/29 30/49 50/64 Male Female Primary Secondary Tertiary Single parent Dual parent No dependent children Rural to large town City or suburb Age Gender Education Household Density Started working from home post-COVID, % Teleworkability and the COVID-19 crisis: a new digital divide? 29 1. task contents of work, which classifies tasks according to the object on which they operate and the type of transformation process involved, and at the highest level it differentiates between a. physical tasks: operating on things, b. information-processing (or intellectual) tasks: operating on information or ideas, c. social interaction tasks: operating on social relations. 2. methods of work, which reflect the forms of work organisation used to coordinate a particular productive process, and which can be broadly characterised by the levels of teamwork, autonomy, and routine in the work process. 3. tools of work, which refer to the technologies used in the production processes. Building on this framework, we define teleworkability as the technical possibility of providing labour input remotely into a given economic process. We say “technical possibility” to emphasize that teleworkability depends on what types of task content can be remotely provided with the available technology. Whether teleworkability – as a potential – is actually put in practice or not in a given work process will also depend on the methods of work (namely work organisation), and the specific tools (technologies) used at work. These are discussed later on in this section. For the moment, we will focus on whether the different types of task content – physical, informationprocessing, and social – are compatible with telework. Information-processing tasks – operating with ideas or information – are arguably the most suited to telework. The Digital Revolution has vastly expanded our ability to transmit, manipulate and store information (Brynjolffson and McAfee, 2017, Fernández-Macías, 2018). Any information that can be digitally encoded without loss can be perfectly transmitted to any part of the world almost instantly, which means that any task aimed at operating with encoded information (verbal, numeric or of any other type) can be performed remotely without any loss in a technical sense. Social tasks can increasingly also be provided remotely, but often with a loss of quality in the service. Digital communications have also expanded massively the possibilities for remote social interaction. Already analogue forms of remote communication such as mail or landline telephones allowed long-distance verbal interaction, but the very significant limitations in the amount of information that could be transmitted (and its speed) made them very inefficient compared to faceto-face communication. In contrast, digitally encoded visual and auditory information transmitted via the internet (as in a videoconference) allows for a very efficient remote social interaction, as long as it is a form of social interaction does not require actual physical contact. There is still significant information loss compared to face-to-face social interaction: even in the most flawless videoconference, there are small lags or informational noise that can be quite disruptive, and also many non-verbal or connotative clues are likely to get lost (for instance, see Schoenenberg et al., 2014). From the taxonomy of social interaction tasks of Fernández-Macías and Bisello (2020), we can list some categories of social tasks that can be remotely performed but with a loss of quality: teaching, selling, negotiating, caring, coordinating. Finally, physical tasks – which involve the physical operation with things or people – are the ones where remote labour input remains most difficult. Although there have been important advances in recent years in the fields of telepresence and telerobotics (for a discussion, see Baldwin, 2019), being able to remotely manipulate objects with a level of precision similar to actual presence is still beyond the capabilities of even the most advanced technologies. Therefore, the physical task content of any occupation will determine whether it can be performed remotely or not. Any job that requires significant physical contact with things or people will not be amenable to telework. Teleworkability and the COVID-19 crisis: a new digital divide? 30 Therefore, following the framework of Fernández-Macías and Bisello (2020), we can differentiate three big categories of tasks by their teleworkability given the state of existing technology: • physical tasks, which can generally not be provided remotely with existing technologies and thus are the real bottleneck for the teleworkability of occupations; • social interaction tasks, which unless they require physical contact can also be provided remotely but with a significant loss of quality; • information-processing tasks, which can in general be provided remotely with hardly any loss and which can be easily identified by their use of computers. As discussed in Fernández-Macías and Bisello (2020), in practice, the vast majority of jobs incorporate a mix of physical, information-processing, and social interaction tasks. In advanced service economies such as those of the EU, the predominant type of task content in the labour market is social interaction, followed by information-processing and last (with the lower prevalence) physical task content. But since physical tasks are the real bottleneck of the teleworkability of occupations, we can operationalise it into a negative and binary indicator of technical teleworkability: if a job has a significant amount of task content that requires the physical manipulation of objects or people, then we can classify it as not teleworkable. Dingel and Neiman (2020) also use a binary and negative approach to measure teleworkability (though including other types of task content, see Box 5 for more details). For the much more common social interaction tasks, we can develop a negative and continuous indicator that complements the index of technical teleworkability. As previously argued, as long as a social interaction task does not require direct physical contact (in which case it would fall in the physical category), it can in principle be carried out remotely but with a decline in quality. For a given job, this decline in quality associated with telework can be approximated by the amount of (non-physical) social interaction required. For instance, a psychotherapist job is almost entirely composed by social interaction task content, but it does not require physical contact: therefore, the job can be classified as teleworkable according to the logic of the previous paragraph, adding the further qualification that in this case, telework implies a significant loss of quality in the provision of the service. To operationalise this qualitative assessment into a continuous indicator for occupations, we can use indicators of social interaction tasks at work (for instance, measuring the share of total working time spent on social tasks) and aggregate them into an index under the assumption that the more social interaction tasks the lower the quality of the service provided if it is via telework. Thus, a value of 0 in this index (meaning no social interaction tasks) would be the most teleworkable, and a value of 1 the least, with values in between reflecting the degree of social interaction tasks present in the job. However, it is important to note that this index may not be interpreted in terms of teleworkability on its own, but only in combination with the physical teleworkability index previously discussed. This is because, as already argued, physical interaction task content is the actual bottleneck, and thus a value of 0 in that index implies absolute nonteleworkability; if a job has a value of 0 in terms of physical teleworkability, the value of social interaction is irrelevant. Only when a job is physically teleworkable can we use the social interaction index to qualify its teleworkability as continuous attribute: a physically teleworkable job with a lot of social interaction is less amenable to remote provision than a physically teleworkable job with no social interaction. Finally, since physical and social task content determine negatively the teleworkability of a given job, the extent to which the same job involves information processing is ultimately irrelevant, even if information processing tasks are generally teleworkable. As we will discuss in the following section, we will only use a proxy indicator of information processing tasks (use of computers at Teleworkability and the COVID-19 crisis: a new digital divide? 31 work) as complementary information, to validate the results of our main variables of teleworkability (based in physical and social interaction task content, as previously explained). One may be tempted to integrate the physical and social indicators into a single index, so that a value of 0 in the resulting composite index reflects no teleworkability (based on the physical content of the job), a value of 1 reflects the highest teleworkability (no physical or social interaction tasks), and values in between reflecting (in reverse) the degree of social interaction tasks for non-physical (and thus technically teleworkable) jobs. However, in practice, it is useful to keep the two indices separate because they convey more information, as illustrated in the next sub-section. Constructing indices of teleworkability To construct the teleworkability indicators of physical and social interaction, we relied on existing European data sources that measure the task content of specific occupations with some level of detail. In particular, we used the Italian Indagine Campionaria delle Professioni, and the European Working Conditions Survey. The Italian Indagine Campionaria delle Professioni (ICP) is a survey of occupations conducted by the National Institute for Public Policy Analysis (INAPP) in collaboration with the Italian National Statistical Institute (ISTAT). It follows closely the structure of the American O*NET database, and thus incorporates very detailed information on tasks, skills, work contexts and organisational characteristics, collected at the 5-digit level of occupations (Codici Professionali, or CP). There are two waves currently available (2007 and 2012), the most recent being the one used in this paper. In total, 16,000 Italian workers were interviewed, representative of sectorial, occupational and geographic heterogeneity (see also Cetrulo et al. 2019). On average, 20 workers per each 5-digit occupation were interviewed face to face for around one hour. The scope and depth of the survey, at the level of detailed occupations, makes it uniquely able to measure in the content of work across occupations. We chose to use this source over the data of American O*Net because we believe it better reflects the characteristics of occupations, tasks, and work organisation found in the European economy. In recent work adapting the Dingel and Neiman approach to the Italian labour market, Cetrulo et al (2020) have made a similar use of the ICP to that in this paper estimating that 30% of Italian workers are employed in teleworkable jobs. The European Working Conditions Survey (EWCS), conducted by Eurofound, focuses on the conditions of work and employment of Europeans. It is representative of the entire EU employed population, and conducted every five years since 1991. The wave we use in this paper (2015) covered 35 countries and nearly 44,000 interviews (between 1,000 and 2,000 interviews per country). It was also conducted face to face, in the homes of the respondents, for an average duration of about 45 minutes. Technical teleworkability index From these surveys, we identified variables consistent with the approach to teleworkability presented in the previous pages. First, for tasks involving the physical manipulation of objects, people, or machinery, we identified the following seven variables: Teleworkability and the COVID-19 crisis: a new digital divide? 32 Table 1: Variables selected for technical teleworkability index Variable Scale reported Unit Source Manual Dexterity Importance (0-100) CP 5 digit ICP Finger Dexterity Importance (0-100) CP 5 digit ICP Performing General Physical Activities Importance (0-100) CP 5 digit ICP Handling and Moving Objects Importance (0-100) CP 5 digit ICP Inspecting Equipment, Structures, or Material Importance (0-100) CP 5 digit ICP Operating Vehicles, Mechanized Devices, or Equipment Importance (0-100) CP 5 digit ICP Lifting or moving people Frequency (7-point scale) ISCO 3 digit EWCS All of these tasks involve the direct and physical operation with things or people, generally requiring strength, dexterity and hand-eye coordination. Although there are experimental technologies that may eventually achieve a breakthrough in the remote provision of this kind of labour input (for instance, remotely controlled robots), as of today these technologies are very far from being widely available. Therefore, these types of task content cannot be performed remotely. Once these variables were identified, we proceeded to standardise and aggregate them.11 In the case of EWCS, the original 7-point measurement scales were translated into a continuous scale of 0-100 (from lowest to highest physical task intensity in the occupation). The data of ICP, originally collected on a 7-point scale in surveys, was already reported on a 0-100 scale. In the case of ICP, for each 5-digit occupation, whenever any indicator of the six physical task variables was above a threshold of 40, we classified the corresponding occupation as not teleworkable.12 This is based on our understanding that if respondents from a given occupation report that even one physical task is important for their job, then the job in question cannot technically be carried out remotely. However, setting a numeric threshold for what “sufficiently important” means is necessarily an arbitrary choice. Based on the distribution across occupations of the raw response values of the importance scores, we decided on a value of 40 on the 100-point scale as a threshold for all variables from ICP and EWCS, with a few minor ad-hoc adjustments commented in greater detail in the Annex. As a result, each of the 798 5-digit occupations in ICP was classified as either technically teleworkable or not. Then, to allow for international comparisons, we aggregated the occupational classification from 798 5-digit Codici professionali (CP) units into 121 3-digit ISCO units, using the official mapping published by Istat.13 Since 3-digit ISCO groups may occasionally bring together occupations with different levels of physical interaction, when aggregating from 5-digit CP into 311 The code for constructing the teleworkability indicators is available at: https://github.com/m-sostero/teleworkoccupations 12 For instance, the occupation 3.1.1.2.0 (Tecnici chimici, belonging to ISCO group 311 Physical and engineering science technicians) received the following scores: 29.8 for manual dexterity tasks, 32.1 for finger dexterity, 15.5 for performing general physical activities, 40.5 for handling and moving objects, 54.8 for inspecting and 10.0 for operation with vehicles or equipment. Since the scores were above the threshold of 40 in “handling and moving objects” and in “inspecting equipment, structures or materials”, the occupation Tecnici chimici was classified as not teleworkable. 13 https://www.istat.it/en/files/2013/07/la_classificazione_delle_professioni.pdf. Teleworkability and the COVID-19 crisis: a new digital divide? 33 digit ISCO we weighted the binary values of the teleworkability index based on the relative share of employment in each 5-digit occupation among the 3-digit group, based on recent Italian Labour Force Statistics. Thus, the value of teleworkability of a 3-digit occupation reflects the employmentweighted share of 5-digit occupations within it that are teleworkable.14 Finally, only for the variable “lifting or moving people” from EWCS, we computed the average value for workers at the level of ISCO 3-digit codes. As for variables drawn from ICP, we classified occupations as not teleworkable if the value was above a threshold of 40. The result is a continuous variable measuring the technical teleworkability of each 3-digit occupational code, ranging from 0 (the occupation is fully non-teleworkable) to 100 (the occupation is fully teleworkable). Values between 0 and 100 correspond to 3-digit occupational categories which incorporate some more detailed 5-digit occupational codes which are teleworkable and some which are not.15 Social interaction index We constructed a continuous index of social interaction task content, which is intended as a qualification of the assessment of technical teleworkability of the previous index. As explained in the previous section, any occupation which is totally or partially teleworkable from a technical perspective can be also be assessed in terms of how socially comfortable and efficient the remote provision of labour will be, as a function of the degree of social interaction involved. For constructing this second indicator, we used the following five variables, all of them from ICP. Table 2: Variables selected for social interaction index Variable Scale Unit Source Selling or Influencing Others Importance (0-100) CP 5 digit ICP Training and Teaching Others Importance (0-100) CP 5 digit ICP Assisting and Caring for Others Importance (0-100) CP 5 digit ICP Performing for or Working Directly with the Public Importance (0-100) CP 5 digit ICP Coordinate the work and tasks of others Importance (0-100) CP 5 digit ICP These five variables cover all of the five main dimensions of social interaction task content of the framework of Fernández-Macías and Bisello (2020). As in the variables measuring physical task content, the five variables were originally measured in a 5-point importance scale, which was converted into a continuous 0-100 scale. Then, for each 3-digit occupation (in this case, there was no need to aggregate at the 5-digit level), we computed the arithmetic average of the two highest 14 Following with the previous example, the final value of teleworkability for ISCO-3 digits code 311 (Physical and engineering science technicians) is 0.7 (in a scale of 0-100), because 13 of the 14 5-digit occupations within 311 are not teleworkable, and the one that is teleworkable (Disegnatori tessili) accounts for less than 1% of employment in that code. 15 For instance, ISCO 3-digit code 221 (Medical doctors) has a value of teleworkability of 39.2. This is because within that code there are six 5-digit occupational codes: general practitioners, specialists, surgeons, pathologists and radiotherapists. While GPs, surgeons, pathologists and radiotherapists are classified as not teleworkable, specialists are teleworkable because none of their physical task indices go beyond 40. Since 39.2% of the 3-digit code 221 corresponds to the teleworkable occupation of medical specialist, that is the value we assign to ISCO code 221. Teleworkability and the COVID-19 crisis: a new digital divide? 34 scores in any of the 5 social interaction variables. The reason for this is that it is impossible for any individual job to involve all the five different types of social task content simultaneously. However, we wanted to assign higher values to occupations that require intense social interaction in two types of social tasks, which is perfectly feasible. Additional indices Finally, we constructed two additional indicators which are only used for secondary purposes (to check the plausibility of the two main indicators, and to provide additional context). These are: • Computer use (from EWCS): a variable which measures the approximate amount of time spent “working with computers, laptops, smartphones, etc”. As with the other variables, the original 7-point frequency scale has been translated into a 0-100 scale reflecting the approximate percentage of time using computers at work, and then average scores by 3-digit ISCO codes have been computed. • Pre-COVID19 work at home (from 2018 European Labour Force Survey): this is an indicator measuring how often the respondent works at home. The original 3-point scale has been rescaled to 0-100, assigning a value of 100 to those that “usually” work at home, 30 to those that “sometimes” work at home, and 0 to those that “never work at home. The scores were finally averaged by 3-digit ISCO codes. Advantages and limitations of the methodology The approach to construct the teleworkability indices described above is intended to make a plausible classification of occupations in terms of whether they can technically be performed remotely. Following the framework and taxonomy of tasks for occupational analysis developed in Fernández-Macías and Bisello (2020), and considering the current state of technology, we argue that the crucial determinant of whether a certain job can be done remotely or not the relative importance of physical tasks – those involving physical interaction with objects or people. However, the same framework also considers work organisation and technology as crucial aspects of the task profile of occupations, and certainly these are aspects that can also affect the actual incidence of telework in particular cases. The methods of work relate to organisational (and legal) factors. The ways in which work is allocated, coordinated and supervised in an organisation surely matter for the existence of telework options. Even enterprises operating in the same sector, or similar public administrations in different countries can organise work in very different ways, which will have an impact on how much work can be done remotely. With the data at our disposal, we cannot easily measure organisation structures and cultures across different types of European companies. Moreover, as the evidence on the current expansion of teleworking presented at the end of the first section shows, work organisation practices and regulation are ultimately contingent on the circumstances, and can change quickly or be suspended in case of necessity. For instance, as previously mentioned, telework is normally much more frequent for managers than for secretaries, but this likely reflects less technical feasibility than hierarchical position. The technical feasibility of remotely providing their typical task content is probably similar for both categories of workers, but managers typically enjoy more autonomy and are subject to less monitoring of their work effort than secretaries. In this paper, we are mostly concerned with estimating the technical feasibility of telework across different occupations. A similar reasoning holds for tools of work, notably the type of information and communication technology (ICT) used at work. The specific technology also certainly matters for telework, as digital communication tools shape different methods of remote work, including collaboration and Teleworkability and the COVID-19 crisis: a new digital divide? 35 teamwork, as well as setting the level of autonomy, latitude or supervision of workers. However, every organisation chooses and uses its technology infrastructure based on its own routines and needs; it is thus difficult to observe how technology affects work organisation in aggregate at the European level. The ICP and EWCS data only describe differences in the use of technology at a superficial level, for instance by focusing on the use of e-mail, a well-established form of communication, which is not particularly informative of teleworkability in itself. Another limitation of the data available is that we cannot account for international differences in the digital skills and use of technologies by different occupations. Whatever detailed data we have on the matter relates to Italy in 2012, which is unlikely to correctly represent the state of technology in European workplaces in 2020. The operationalisation of the technical teleworkability index inevitably involves arbitrary methodological choices. In particular, for analytical simplicity we decided to implement technical teleworkability as a binary indicator (originally defined for CP 5-digit occupations). This implies setting a hard threshold on the values of the underlying physical task indicators reported in the ICP survey, which we set at 40 on a 100-point scale. In the majority of cases, the resulting classification, reported in Annex 2, is consistent with intuition and anecdotal evidence on whether different occupations have been able to telework during the lockdown. Unsurprisingly, there are also some exceptions of occupations that are classified as teleworkable when experience suggests they should no bet, or vice-versa. This is especially the case when aggregating the scores from the original CP 5-digit occupations that the ICP survey measures, to more diverse ISCO 3-digit occupation groups used for further analysis. Upon closer inspection, many of these classifications follow quite consistently from the values of the physical task indicators of the underlying occupations. Partially teleworkable ISCO 3-digit groups, with values of technical teleworkability between 0 and 1, reflect this heterogeneous composition of some 5-digit occupations that can telework, while others cannot, at least in terms of their relative weights in the Italian labour force. In two cases, we decided to re-classify some occupations, on the ground that the physical interaction variables reported in the survey fell slightly below the threshold, but did not seem to accurately reflect the task profile of those occupations. This was notably the case of cashiers, which reported (implausibly, in our view) to have limited physical interaction, including with objects, in their job. We also reclassified child care and teacher’s aides as not teleworkable, on the same grounds. We comment the classification and discuss all such cases in Annex 1. However, these ad-hoc adjustments only concern a tiny minority (2) of the occupations classified (798 at 5 digits, 121 at 3). In general, we should note that our classification methodology relies on a snapshot of the nature of work taken before the COVID crisis and lockdown, and is especially reliant on detailed data from a single country. In general, the tasks, methods and tools of work varies across countries and changes over time, although existing evidence suggests that the cross-country variation in task content (the core of our classification) is much smaller than in task methods and tools (Fernández-Macías, Hurley and Bisello 2016). The unprecedented circumstances and exigencies of the lockdown have probably accelerated these changes, forcing companies and workers to adapt. Some forms of work organisation practices that were rooted in office work have probably been suspended, at least temporarily. The tools of work, notably corporate ICT equipment, have also been adapted or expanded to enable mass remote connection. It is also very likely that workers, making a virtue of necessity, had to change the composition and relative importance of tasks they normally perform in their occupations to adapt to the circumstances. No doubt a minority of resourceful people in resilient organisations were able to change the way they work radically, but existing evidence suggests that among the three axes describing work, the task contents of work are more constant in the short term, while methods and tools are more contingent or flexible (Fernández-Macías, Hurley and Bisello, 2016). Teleworkability and the COVID-19 crisis: a new digital divide? 36 Box 5: Comparing the methodology of teleworkability with Dingel and Neiman (2020) Dingel and Neiman (2020) provided estimates of the share of the US population that can work from home, based on data from the American occupational database O*NET. From this survey, which collects detailed standardised information on what workers do in different occupations and is the model of the Indagine Campionaria delle Professioni used in this work, they identified a number of variables to infer whether the associated occupations could be performed from home or not, from the sections on “Work context” and “Generalised work activities”. The most important difference between our approach to identify the teleworkability of occupations and that of Dingel and Neiman is the theoretical justification. Whereas we try to anchor our teleworkability measure to a tasks framework developed for occupational analysis, and provide a detailed justification for the items included and the operationalisation, Dingel and Neiman do not provide much justification for the variables they pick and the method of coding and aggregating them. In our view, an explicit theoretical anchor and justification helps to link our findings with the previous literature and to make a better interpretation and discussion of the broad implications. As in our technical teleworkability index, Dingel and Neiman define teleworkability in general in a negative and binary way: some specific values of the identified variables are considered as incompatible with telework, and thus they are used to code the associated occupation as nonteleworkable. For instance, if the average worker in a given occupation uses email less than once a month, the occupation is coded as non-teleworkable. However, Dingel and Neiman do not provide a general logic or a theory for the selection of variables or the conditions that define teleworkability: in a methodological annex, they just list the O*NET variables and thresholds used. In total, they used 15 variables measuring very different aspects of work – negative indicators of teleworkability as previously mentioned – which we can classify (our criteria, not theirs) in four groups: • Physical work content: performing general physical activities, walking or running, handling or moving objects. Here we can also include a series of variables that refer to the handling and operation of machinery: controlling machines and processes, operating vehicles and equipment, repairing and maintaining mechanical or electrical equipment, and inspecting equipment, structures or materials. • External conditions of work: work outdoors, wearing protective equipment, exposure to burns or bites, exposure to disease or infection. • Social interaction at work: performing or working directly with the public, dealing with violent people. • Use of communication technology: use of email. Looking at the list above, we can infer some criteria implicitly used by Dingel and Neiman for identifying the teleworkability of occupations. According to the number of related O*NET variables used (8), it seems that performing a significant amount of physical or manual activity is the most important criteria to determine that an occupation cannot be performed remotely. A second criterion would relate to physical exposure to external conditions at work (3 variables); a third to face to face social interaction (2) and a final criterion would be the use (or lack of it) of a specific communication technology – email. Teleworkability and the COVID-19 crisis: a new digital divide? 37 As previously argued, this seems reasonable and may work as a practical approximation to identify the teleworkability of occupations, but it also seems somewhat ad-hoc, and is not very clearly justified. For instance, why using only one variable related to the use of communication devices, and why choose email, when telephone is also considered in O*NET? Why include a variable specifically about dealing with violent people, when that can surely be also done (and more comfortably so) by phone or other remote means of communication? Why include so many variables measuring operation with machines? What is the underlying theory or conceptual framework for characterising the teleworkability of occupations? In general, they include many variables specifically related to the (manual) operation with machinery, which in our view can be problematic: although it is probably true that nowadays most machine operation is still carried out manually, an increasing amount of machinery incorporates digital connectivity and sensors which are specifically designed to facilitate remote operation. Dingel and Neiman also use variables measuring external conditions of work (work outdoors, wearing protective equipment, etc.), which may work well as a proxy for physical task content – most work outdoors is of a physical nature, in most cases because it implies interaction with the physical environment, as in gardening or construction – but in itself seems unrelated to teleworkability. Although Dingel and Neiman also take social interaction into account, in this respect our approaches differ more significantly: in this case, they also apply a negative and binary approach that in our view does not quite fit with social interaction. As previously argued, we believe social interaction affects the quality rather than the strict feasibility of remote work. They only use two indicators of social interaction tasks, one of which does not even fit in our view (dealing with violent people, which is not something that can only be done face to face). Finally, Dingel and Neiman include one indicator of use of email at work that seems to imply a different approach to teleworkability – based on technology use rather than task content – which we avoid because in our view ICT use is just too pervasive to be used for identifying teleworkable jobs. Other recent papers approach this issue in slightly different ways. Leibovici et al. (2020) focus only on physical proximity to people: based on O*NET variables that assess the extent to which the job requires workers to perform tasks in close proximity to other people, they classify occupations into low, medium and high contact-intensity occupations. Mongey et al. (2020) combine the two previous approaches by Dingel and Neiman (2020) and Leibovici et al. (2020) to construct two indicators of teleworkability: “work from home” and “physical proximity”. Their analysis shows a strong negative correlation between these two indicators. However, this also highlights some important outliers, such as education jobs which require high physical-proximity but exhibit few features that would prevent working from home, suggesting that under strict confinement rules these jobs would be teleworkable but would have to adjust to comply with the social distancing measures. Pouliakas and Branka (2020) developed a COVID-19 social distancing risk index based on physical proximity and digital intensity to identify individual and job factors most likely to be impacted by social distancing measures, and found that a majority of the EU-27 workforce is at varying degrees of risk of disruption. Teleworkability and the COVID-19 crisis: a new digital divide? 38 What jobs are teleworkable? This section presents first some descriptive analysis of the teleworkability and social interaction indices for ISCO 3-digit occupation groups and their correlations with computer use and previous experience with telework. It then presents EU-level estimates on how many and what types of jobs are teleworkable in Europe, based on these indicators and 2018 EU Labour force Survey Estimates. Teleworkability indices of occupations Figure 18 shows the values of technical teleworkability and social interaction indices for the 121 ISCO 3-digit occupation groups. The figure is divided in three panels: The panel on the left-hand side shows the values of social interaction for all the occupation groups that were classified as fully non-teleworkable. These ISCO 3-digit occupation groups are entirely composed of CP 5-digit occupations that involve a significant amount of physical task content (above 40 in at least one of the 7 specific task indicators used), and thus we consider that they cannot be carried out remotely. Most of these occupations have relatively low values of social interaction (the points cluster at values below 50 in the vertical axis), although there are some exceptions. Nurses and midwifes, street and market salespersons, childcare workers and street vendors are examples of occupations that are not teleworkable but involve very high levels (around or above 75 points in our index) of social interaction. But at the bottom half of the panel we see the bulk of non-teleworkable occupations, including most manual occupations in manufacturing, transport and mining. These are occupations that cannot be teleworked because they incorporate a very significant amount of manual labour, but they involve very little social interaction task content (beyond interactions with co-workers). The panel in the middle of the chart shows the values of technical teleworkability and social interaction for occupations that are partially teleworkable, meaning that the ISCO 3-digit group encompasses some CP 5-digit occupations that are teleworkable and some that are not. As already implied in the left-hand panel, there is some positive correlation between the technical teleworkability and social interaction indices, because occupations with high teleworkability tend to involve more social interaction (the points tend to move from the bottom left to top right quadrant). Occupations with low teleworkability and low social interaction are machine operators, control technicians and elementary workers. On the opposite quadrant (high teleworkability, high social interaction) we find teachers, doctors, travel attendants, managers in services. There is also an interesting cluster of occupations with high teleworkability and low social interaction, such as clerical support workers, life science technicians and ICT operators. Finally, the panel on the right-hand side shows the social interaction scores of all the occupation groups classified as fully teleworkable. Again, the positive association between technical teleworkability and social interaction is confirmed, because there are more occupations above than below the middle value of social interaction (50). At the top, we see fully teleworkable occupations that involve a lot of social interaction, including teachers, legal associate professionals, managing directors, legislators, sales agents and brokers. Below a score of 50 in social interaction task content, we see many clerical occupations, some ICT and technical service professionals, authors, journalists and linguists. Thus, Figure 18 serves as a good summary of our assessment of teleworkability across detailed occupational categories in Europe, based in a framework that focuses on the physicality of jobs and the degree of social interaction. Since the horizontal axis reflects physical teleworkability and the vertical axis reflects social interaction, we can also analyse the figure by splitting the points in four quadrants: • The bottom right quadrant would include the most easily teleworkable jobs, because they involve very little physical task content and very little social interaction. Examples of Teleworkability and the COVID-19 crisis: a new digital divide? 45 Teleworkability across the EU With the occupational teleworkability indicator developed in the previous pages, we can now look at what this means in terms of employment share by member state and for different categories of workers. The question we are trying to answer is what share of current dependent employment is working in occupations that can technically be carried out at home, meaning that there are no specific physical requirements of the job that would prevent it. We derive these estimates by simply multiplying the technical teleworkability scores of ISCO 3-digit occupation groups, derived in the previous section, by their employment levels across the EU as measured by the 2018 Labour Force Survey. For those 3-digit ISCO occupations groups that are only partially teleworkable – those with a technical teleworkability score between 0 and 1 – this projection implies that only a corresponding share of those employed in the group are considered teleworkable. The first observation from these estimates is that the share of employment that is potentially teleworkable (around 37%) is much greater than the actual prevalence of teleworking pre-outbreak (15%) among all EU27 workers. This gap is equivalent to over one in five workers (22%, or 43 million workers) in the EU27 who could be working from home but did not do so before the COVID19 crisis. The second main observation is that nearly all this gap is comes from the ranks of dependent employees (see Figure 21). Figure 21: Share of teleworkable employment compared to incidence of teleworking (2018), EU27 Source: LFS. Note: ‘teleworkable’ refers to share of employment in teleworkable occupations according to our operationalisation; ‘works from home usually or sometimes’ refers to share of employment from LFS 2018 microdata (EU27) For the self-employed – who in principle have much greater discretion over how and where their work is carried out – there is a much closer correspondence between actual teleworking and technical teleworkability. Indeed, a somewhat higher share of the self-employed reported working from home based on LFS 2018 (34%) compared to their share of employment in teleworkable 0 5 10 15 20 25 30 35 40 Self employed Employee Teleworkable Works from home usually or sometimes Teleworkability and the COVID-19 crisis: a new digital divide? 46 occupations (32%).16 A higher share of dependent employees than of the self-employed is working in technically teleworkable occupations, so if we assume that during the COVID-19 crisis everyone who can telework is teleworking, this would imply a reversal of the actual pre-outbreak prevalence. Henceforward, we focus on the category of dependent employees, which accounts for over 5 in 6 workers across the EU27, and where the gap between potential and actual teleworking is greatest. A first striking difference in teleworkability relates to the white-collar / blue-collar occupational divide (see Figure 22). White-collar work is much more teleworkable than blue-collar work, where the physical requirements of the jobs and associated place-dependence render most occupational categories non-teleworkable. Figure 22: Teleworkability and actual teleworking among employees by broad occupation group Source: LFS, COVID group. Note: employees only. “Teleworkability” refers to share of employment in teleworkable occupations according to our operationalisation; ‘Current telework’ refers to share of employment working from home usually or sometime according to LFS 2018 microdata (EU27). This differential was also observed for teleworkable employment in Italy in separate analysis (Cetrulo et al, 2020). In the ISCO occupational groups 7-9 (craft and related trades workers, plant and machine operators and elementary occupations), the share of teleworkable employment is less than 2%; and actual incidence of teleworking is 1% or less. By contrast, in ISCO groups 1-4, a large majority of employment is technically teleworkable (from 54% of associate professional employment to 85% of clerical support workers). While this correlates to a higher incidence of actual teleworking in these occupations compared to the blue-collar occupations, the gap remains large between actual and potential teleworking. While 16 Due in part to our permissive operationalisation of actual homeworking as those who indicate either working from home ‘usually’ or ‘sometimes’, which may lead to over-estimation. 0 10 20 30 40 50 60 70 80 90 Share of employment (%) Teleworkability % Current teleworking % Teleworkability and the COVID-19 crisis: a new digital divide? 47 71% of professionals could work from home, around a quarter (24%) actually did so. The gap is even wider in the case of lower-level white collar occupations (clerical support workers) where the characteristics of the job are such that almost all could be teleworked (84%) while actual teleworking incidence remains very marginal (5%). This supports our hypothesis that access to teleworking has hitherto been conditioned by position in the occupational hierarchy and associated privileges, more than by the task composition of the job. The evidence in this report suggests that during the COVID-19 crisis the large expansion of telework has followed the occupational distribution indicated below – a strong skew towards white-collar employment where much work has shifted from office to home, while workplace closures have resulted in furloughing (or job loss) for blue-collar occupations where there is little realistic option of teleworking. Over and above this, it is likely that access to telework has become more evenly distributed within white collar occupations on the assumption that enforced closures have resulted in many new teleworkers amongst low and mid-level clerical and administrative workers who previously had limited access. The teleworkable employment shares are predictably higher among service sectors with high shares of white-collar employees (see Figure 23). Figure 23: Teleworkability in EU27 by sector Source: LFS, COVID group. Note: employees only Nearly all financial-services employment is teleworkable (93%) as well as nearly four in five employees in information / communication (79%) and around two-thirds of employees in real estate, professional, scientific and technical activities, education and public administration. Service sectors with lower shares of teleworkable employment include health (30%), retail (27%) and accommodation/food services (16%). The primary sector, manufacturing and construction sector all have low shares of teleworkable employment (10-20%). 0 10 20 30 40 50 60 70 80 90 100 Financial and insurance activities Extraterritorial organisations Information and communication Education Professional, scientific and tech activities Real estate activities Public administration and defence Arts, entertainment and recreation Other service activities Electricity, gas, steam supply etc Admin and support service activities Transportation and storage Human health and social work activities Wholesale and retail trade Water supply, sewerage, waste mgmt Manufacturing Accommodation and food services Mining and quarrying Construction Agriculture, forestry and fishing Activities of households as employers Teleworkable employment, % of dependent employment Teleworkability and the COVID-19 crisis: a new digital divide? 48 Because we measure teleworkability as an attribute of occupations, constant for the same occupations across different countries, the share of teleworkable employment varies by member states based only on the composition of their workforce (see Figure 24). Figure 24: Share of teleworkable employment, by country, in EU27 Source: LFS, COVID group. Note: employees only This means that countries with higher shares of white-collar occupations – and the sectors where these are more prevalent – have a larger share of employment that is teleworkable. The estimates thus cannot take into account differences between member states in ICT technology, regulation, and size of companies. According to our estimation, the range of teleworkable employment is from 27% in Romania to just over twice that in Luxembourg (54%). The highest share of teleworkable employment tends to be in the Nordic and Benelux countries, and correlates largely with the ranking of countries by preoutbreak teleworking prevalence. The lowest shares are in Eastern Europe and also in some of the larger member states in Southern Europe (eg., Italy and Spain) where, as we have already seen, other sources point to relatively low teleworking incidence. Overall, there is relatively limited variation between countries: in all member states except five, the share of teleworkable employment ranges between 33 and 44% of total employment. Finally, Figure 25 (next page) summarises differences in the teleworkable share of employment by various personal, job-related or spatial characteristics. 0 10 20 30 40 50 60 RO Romania SK Slovak Republic BG Bulgaria HU Hungary CZ Czech Republic ES Spain HR Croatia LV Latvia PT Portugal IT Italy PL Poland EE Estonia EU LT Lithuania AT Austria IE Ireland SI Slovenia FI Finland DE Germany DK Denmark FR France GR Greece CY Cyprus MT Malta SE Sweden BE Belgium NL Netherlands LU Luxembourg Teleworkability and the COVID-19 crisis: a new digital divide? 49 Figure 25: Share of employees in teleworkable occupations Source: LFS, COVID group, Structure of Earnings Survey. Note: employees only. Job-wage quintiles based on author’s calculations of SES 2014 data. The most obvious determinants of whether an occupation is teleworkable or not is the wage and education level of the average job-holder. Three-quarters (74%) of those in jobs in the highestpaying quintile can telework. The teleworkable shares of employment are well above average also in mid-paid and mid-high paid jobs (Q3-Q4) before falling off sharply at the bottom of the job0 10 20 30 40 50 60 70 80 Full-time job Part-time job Permanent Temporary Adult(s), no dependent children Single parent Two adults, dependent children 15 to 29 30 to 49 50 to 64 65+ Native-born Foreign-born Micro <=10 Small 11-49 50+ Cities Towns and suburbs Rural areas Male Female Low: Lower secondary Medium: Upper secondary High: Third level Q1 - low paid Q2 Q3 Q4 Q5 - high paid Full / part-time Contract type Household composition Age Country of birth Establishment size Settlement type Gender Education Wage quintile Teleworkable employment, % dependent employment Teleworkability and the COVID-19 crisis: a new digital divide? 50 wage distribution. Only 3% of those in the lowest quintile are in teleworkable employment. Given the strong correlation between job-wage and job-qualifications, it is not surprising to see this sharp wage gradient of teleworkability reflected in the differentiation by educational qualification. Around two-thirds (64%) of third-level graduates work in teleworkable occupations, but a much smaller share of those with lower levels of qualification. Box 6: teleworkability, social interaction and computer use In our approach, the sole determinant of occupational teleworkability in is the lack of physical handling tasks. As indicated earlier, the quality of teleworkability will depend also on the extent to which an occupation requires extensive social interaction. Where an occupation is rich in social interaction tasks, telework will inevitably involve some loss of quality given the limitations of existing ICT technologies. Where an occupation is teleworkable and involves only limited social interaction, it should in principle be capable of being carried out with no or limited loss of quality. Prospectively, it is in these latter occupations where, if there is a general increase in teleworking post-COVID, it could be expected to occur earliest and fastest. Conversely, in occupations such as those of secondary school teachers, the expectation would be that post-COVID, most work while technically teleworkable would revert to its traditional locus in schools and classrooms. Table 3: three categories of teleworkability*, characteristics, EU27 Occupation Type Employment share % Computer use Social interaction PreCOVID telework Physical teleworkability Occupation (n, ISCO3d) Example Not teleworkable 66 27 47 5 7 83 Doctors Telework, limited social interaction 13 86 41 11 97 14 Finance professionals Telework, extensive social interaction 22 67 65 17 90 26 Secondary School teachers Source: LFS, EWCS, ICP. Teleworkable occupations account for 37% of dependent employment, but the majority of teleworkable jobs are those with extensive social interaction where teleworking may not be optimal. This was also the case in terms of pre-COVID telework incidence. Only 13% of employment is in occupations that are teleworkable and involve limited social interaction. In these occupations (eg., finance professionals) computer use is more intensive than in teleworkable jobs with extensive social interaction (eg., secondary school teachers) and much more intensive than in non-teleworkable occupations. In other words, beyond the main physical teleworkability criterion, computer use intensity is a strong additional or complementary indicator of teleworkability. Note *: three occupational categories generated as follows. A cut-off point of 40 on the physical teleworkability scale to differentiate between teleworkable and non-teleworkable occupations (ISCO 3d) – as in the construction of the original index itself. For this reason, the shares of employment diverge marginally from those already presented in this section and the physical teleworkability share is positive and non-zero (7%) in the ‘not teleworkable’ category. We then further differentiate teleworkable jobs based on the social interaction scale discussed in the previous section and use a cut-off to differentiate between occupations with limited (<50) or extensive (>=50) social interaction. Computer use intensity = occupational mean scores based on 6EWCS. Pre-COVID telework based on LFS 2018 ‘homework’ question scaled as follows – sometimes=0.4, usually=1, never=0. ISCO categories excluded: armed forces, subsistence farmers. Teleworkability and the COVID-19 crisis: a new digital divide? 51 Another striking finding from Figure 24 (page 48) is that a much higher share of women than men (45% compared to 30%) are in teleworkable occupations. As we have seen, the incidence of teleworking has been similar by gender in pre-outbreak representative surveys, though the Eurofound COVID-19 survey data did show a small differential (41% of female employees and 37% of male report starting to telework since the onset of the current crisis). The gender difference in teleworkability relates in part to patterns of sectoral segregation with men in particular overrepresented in sectors with limited teleworkability potential such as agriculture, mining, manufacturing, utilities and construction. Even in these male-dominated sectors, the female teleworkable share of employment tends to be high. In construction, only 6% of male employment is teleworkable compared to 69% of female employment with similar differentials observed in utilities, mining and transportation and storage. This reflects the fact that female workers tend to work in different jobs than men in these sectors and that these jobs tend to be the teleworkable ones – office-based, secretarial or administrative in nature, with a lower share of physical handling tasks. In addition, within the main occupational groups, the female teleworkable share of employment tends to be higher; 86% of female managers and 89% of female clerical support workers are in teleworkable jobs compared to 75% and 73% respectively of their male counterparts. Teleworkable employment tends to be more common among native-born workers than foreignborn, among those on permanent contracts compared to those in temporary contracts and among those in larger establishments than smaller establishments. 17 The spatial distribution of employment is also a discriminating factor as cities, with their higher share of services employment – in particular of the knowledge-worker or symbolic analyst type of occupation – have more teleworkable employment (44%) than towns or suburbs (35%), or rural areas (29%). Other characteristics, such as age, full-time / part-time status and the household composition of job-holders, do not appear to differ significantly between teleworkable and non-teleworkable occupations. Overall, based on employment shares from the EU Labour Force Survey and our detailed technical teleworkability index, we find that the distribution of teleworkable employment is consistent along a number of dimensions with the evidence from both preand post-outbreak surveys. There is a large gap between the share of dependent employees who could perform their work remotely and the share that regularly or occasionally did so pre-crisis. We estimate that this gap amounts to over 20% of total EU27 employment or over 40 million workers. This would put our estimate of teleworkable employment (36%) – or 37% of dependent employment – close to the middle of the range of estimates from real-time surveys conducted during the COVID-crisis. The incidence of actual teleworking and technical teleworkability have therefore converged closely because of the crisis. Teleworking is feasible almost exclusively for those in white-collar occupations and largely nonfeasible for those in blue-collar occupations. This skew towards higher-level occupations is starkly evident in a steep gradient of teleworkability by completed educational level and, especially, jobwage quintile with three-quarters of the highest-earning 20% of workers able to work from home but less than one in twenty of the lowest-earning 20%. The striking result related to the gender of teleworkable occupations with females 50% more likely than men to be employed in such occupations – though access to teleworking post-crisis is likely to be inflected by issues of organisational position and power, as was the case pre-crisis, which could tend to narrow or eliminate this gap. 17 Though the EU-LFS categorisation is limited in capturing larger establishments as the largest category is establishments with 50 or more employees. Teleworkability and the COVID-19 crisis: a new digital divide? 52 Discussion Teleworkability and work organisation In this paper, we have defined teleworkability in terms of technical feasibility, which we have argued depends essentially on the technologies available for remote communication and their interaction with the different types of task content (physical, intellectual or social). But the tasks framework that we use in this paper includes as a crucial dimension also work organisation, which refers to how work is socially coordinated. Although we do not focus on this aspect, there are a few considerations that are still useful to make. We will again follow the structure of the framework proposed by Fernández-Macías and Bisello (2020), and thus discuss three main aspects of work organisation: autonomy (and control), teamwork and routinisation. As we have already mentioned, the problem of autonomy and control in the case of telework is probably the main reason why in the last few years it has not become more frequent despite the growing – though not yet universal – availability of ICT. Depending on the implementation, telework can imply a very significant level of autonomy for the worker: when work is carried out from the private home of the worker, the traditional office-based mechanisms to monitor working time and quality of work may not be available, especially for those jobs where output quality and quantity is intrinsically more difficult to measure. In other words, whereas the physical presence of the worker in the employers´ premises facilitates the monitoring and control of those aspects of work (methods, hours, effort), in the case of telework the monitoring of work is much more difficult, and thus requires a much higher level of trust. That is probably why workers in a supervisory capacity (managers) or whose work effort is intrinsically difficult to monitor anyway (professionals) were the most frequently allowed to telework before the COVID-19 crisis. Or to put it the other way round: this is one of the reasons why many mid-skilled workers whose tasks were easily teleworkable with current ICT tools – for instance, clerical workers who mostly perform information-processing tasks – teleworked only very rarely before the COVID-19 crisis, as shown in the previous sections. There may also be cultural norms at play: for instance, work done remotely may be perceived as less visible or recognised within the organisation, which may deter some employees from teleworking, if they fear that doing so would undermine their chance for professional advancement. Clearly, this situation drastically changed with the COVID-19 crisis. Given the public health risks associated with commuting and work presence in this context, all of a sudden teleworkability as a technical feasibility became the sole factor determining who could work from home. Essentially, everyone that could telework had to telework. As entire organisations moved to telework at the same time, under difficult circumstances, even of those who already had some experience of telework may have experienced changes in their routines and work organisation practices. The change was even more radical for the many midand low-skilled workers who could very rarely telework before. Organisations may feel the need to adapt to this arrangement, and may try to introduce increasingly intrusive forms of remote control and surveillance of the work carried out from home. Some recent anecdotal or journalistic evidence on the increasing use of technologies of remote surveillance for employees (for instance, with algorithmic control of the physical presence of the employee in front of the computer as recorded by the camera) points in this direction.18 18 There is no scientific evidence yet on this because it is simply too early, but journalists have been reporting on increasing levels of monitoring work at home from the beginning of the pandemic, see for instance: “Working from home surveillance software for your boss” (Washington Post, April 11 2020), “How My Boss Monitors Me While I Work From Home” (New York Times, May 6 2020), “Employee monitoring software surges as companies send staff home” (ABC News, May 21 2020). Teleworkability and the COVID-19 crisis: a new digital divide? 53 The second aspect of work organisation that is relevant for telework is teamwork and in general social interaction with colleagues. In this respect, we can apply by analogy many of the considerations previously made about social interaction task content: to the extent that a given occupation or job necessarily requires a lot of teamwork or social interaction with colleagues, telework can negatively affect the quality of the labour input provided. We would argue that, since telephone, videoconference and other online collaboration tools are easily available and increasingly of good quality, the need to socially interact with colleagues cannot be considered as a hard barrier to telework, but as something that can make it more or less difficult for different occupations. However, work organisation has more plasticity than task content (which is directly determined by the material properties of the transformation process in each job) and can be more easily adapted to changing conditions if necessary. The third aspect of work organisation to be considered is the level of routine/standardisation of tasks. The routinisation and standardisation of tasks can be used as a form of coordinating work by procedures rather than by direct human intervention, and thus can also facilitate the remote provision of work (by reducing the need of direct coordination of work). In fact, Pouliakas and Branka (2020) show that the task profile of EU remote workers is mostly characterised by standardised tasks. That largely reflects the fact that tasks standardisation, especially when combined with algorithmic control and digital monitoring devices, can actually enable a high degree of control of work processes even when they take place remotely. This is probably one of the key sources of success of digital labour platforms such as Uber or ClickWorker: they are very efficient at coordinating the remote provision of labour at large volumes, solving the autonomy and control problem by standardising tasks, monitoring effort and managing algorithmically. Therefore, although these three aspects of work organisation are not determinants of teleworkability understood as a technical feasibility, they can nevertheless facilitate it or impede it. Therefore, a fuller understanding of the determinants of telework would go beyond technical feasibility and social interaction, and also capture the organisational difficulty of telework in specific occupations, measuring the degree of autonomy (which correlates positively with telework), teamwork (less teamwork, easier to organise telework) and routineness (more routine tasks, easier to organise telework). All these factors would require additional surveys and data to be implemented consistently across the European Union. Some considerations on telework after the lockdown The massive expansion of telework during the COVID-19 pandemic can be considered a large-scale natural experiment in work organisation; an improvised and ad-hoc response at the level of states and organisations to a rare circumstance – a global pandemic, and state restrictions on mobility and physical distancing designed to contain it. Instead of being compelled to work at their employer’s premises, many employees were, for a limited period, compelled to work from home. As our societies gradually return to normality, many employees will resume their routine commute to and from work and the incidence of “full-time” teleworking will once again decline. There are however many reasons to believe that there will be a lasting legacy from the current episode where a quarter or more of employees have been working remotely from home. Some employers are considering changes to their working arrangements, extending the possibility to telework, and may even rethink office-based work altogether. Policymakers have the opportunity to reflect on what the current experience of teleworking means about the nature of work and its future. At least for the next few months, as confinement measures are relaxed, some physical distancing measures are likely to remain in place – limiting the possibilities for a full return to work for many. Rotating telework will be one option with staff rostered to work some days at work, some days at home. Possible fresh outbreaks mean that any resumption of workplace activity is likely to be Teleworkability and the COVID-19 crisis: a new digital divide? 54 tentative and conditional in an initial period lasting many months. For this reason, in the short-term there will continue to be extensive recourse to telework. In a longer-term perspective, after the COVID-19 crisis is resolved by the discovery of vaccinations or other medical remedies, there is likely to be greater acceptance and encouragement of teleworking possibilities based on the knowledge that organisations, HR personnel and managers as well as individual workers have gained from the period of enforced telework. There are also other factors likely to increase the incidence of teleworking in the long term. Some are structural, some employer-motivated, some employee-motivated. On the structural side, compositional changes in employment – notably if the current shift to services, increases in educational attainment and skills, and overall occupational upgrading continue – mean that a higher share of employees will be doing work that is amenable to teleworking. As this paper has demonstrated, such work is heavily skewed to higher-level white-collar occupations. At the same time, the continual upgrading of IT infrastructure, including high-speed broadband availability and enhanced online collaboration tools – in videoconferencing and team collaboration for example – will enable virtual work connections for more employees regardless of location and reduce the ‘quality gap’ between remote and in situ work. Employers may have diverse reasons for wanting to broaden access to telework. Some of the leading technology firms have already indicated an openness to or encouragement of much more extensive teleworking in the near future. If trends in Silicon Valley are any guide, Facebook chief executive, Mark Zuckerberg, has predicted that half of the company’s employees may work permanently from home by 2020 (Sandler, 2020). His counterpart at Twitter has said that postoutbreak all staff wishing to continue working remotely would be allowed to do so ‘forever’ and granted extra company money to fund the necessary physical equipment. In part, this responds to a perceived demand both among current and potential recruits; organisations that do not offer this possibility may in the future be considered less attractive places to work, a trend that was already apparent in recent years (Mandl et al., 2015). The period following COVID-19 outbreak has provided employers with an unplanned organisational learning opportunity in telework provision from the perspective of human resources, work organisation, productivity and ICT infrastructure and software. Some employers will have learnt that broader access to teleworking was technically achievable with limited negative impacts on performance. This learning-by-doing is likely to have persuaded previously sceptical managers and employers that a broader application of teleworking, beyond the happy few, is both feasible and desirable. The lessons learnt will facilitate contingency planning of organisations in a broader sense; work that is place-independent is intrinsically more resilient. Employers are likely also to be motivated by cost considerations, for example in relation to building and office rental and maintenance. The prospect of greater shares of employees working remotely (at least part of the time) may lessen the reliance on rented space. For cash-strapped businesses post-outbreak, this is likely to be an important consideration. In tandem, more remote working may also encourage a related cost saving in reduced work travel as some events, meetings and staff training sessions move online. An important caveat is that employee preferences indicated in surveys are for occasional telework – i.e., a mix of workplace presence and working from home on alternate days. Real estate cost savings for employers would be harder to achieve in this scenario without extensive ‘hotdesking’ which itself may entail additional management and administration overhead. Employers will also need to confront new challenges posed by a change in working arrangements, such as dealing with employee’s work-related health and safety outside the office, monitoring performance and working time, and effectively coordinating workforce remotely. There may also be a need to compensate for the expenses incurred by employees to set up workstations and internet connections, and associated energy expenses. Many of these issues call for discussion between social partners and may require new legislation. Teleworkability and the COVID-19 crisis: a new digital divide? 61 Fernández-Macías, E., Hurley, J. and Bisello, M. (2016), What do Europeans do at work? A task-based analysis (European Jobs Monitor 2016), Publications Office of the European Union, Luxembourg. Fernández-Macías, E. (2018). Automation, digitalisation and platforms: Implications for work and employment. Eurofound Working Paper. Fernández-Macías, E., Klenert, D., Antón, J. (2020). Not so disruptive yet? Characteristics, distribution and determinants of robots in Europe. JRC Working Papers on Labour, Education and Technology, No. 2020/01. Seville: Joint Research Centre, European Commission. Frey, C. B., & Osborne, M. A. (2017). The future of employment: How susceptible are jobs to computerisation?. Technological forecasting and social change, 114, 254-280. Glaeser, E. (2011). Cities, productivity, and quality of life. Science, 333(6042), 592-594. Gottlieb, C and J Grobovsek and M Poschke (2020) “Working from Home Across Countries,” Covid Economics: Vetted and Real-Time Papers 8, 22 April. Guyot, K. and Sawhill, I. (2020), “Telecommuting will likely continue long after the pandemic”, Brookings Institution web posting (dated: 6/4/2020). Link: https://www.brookings.edu/blog/upfront/2020/04/06/telecommuting-will-likely-continue-long-after-the-pandemic/ Hurley, J., Fernández-Macías, E., Bisello, M., Vacas‑Soriano, C. and Fana, M. (2019), European Jobs Monitor 2019: Shifts in the employment structure at regional level, European Jobs Monitor series, Publications Office of the European Union, Luxembourg. ILO (2020) “Defining and measuring remote work, telework, work at home and home-based work”. ILO policy brief. Mascherini, M. and Bisello, M. (2020), Covid-19 fallout takes higher toll on women, Social Europe (23/6/20). Link: https://www.socialeurope.eu/covid-19-fallout-takes-higher-toll-on-women McCarthy, A., Ahearne, A., Bohle Carbonell, K., Ó Síocháin, T. and Frost, D. (2020). Remote Working During COVID-19: Ireland’s National Survey Initial Report. Galway, Ireland: NUI Galway Whitaker Institute & Western Development Commission. Mandl, I., Curtarelli, M., Riso, S., Vargas, O. and Gerogiannis, E. (2015), New forms of employment, Publications Office of the European Union, Luxembourg Messenger, J., Vargas Llave, O., Gschwind, L., Boehmer, S., Vermeylen, G. and Wilkens, M. (2017), Working anytime, anywhere: The effects on the world of work, Publications Office of the European Union, Luxembourg, and the International Labour Office, Geneva. http://eurofound.link/ef1658 Messenger, J. (ed) (2019), Telework in the 21st century: an evolutionary perspective, ILO / Edward Elgar. Link: https://www.ilo.org/global/publications/books/WCMS_723395/lang--en/index.htm Möhring, K., Naumann, E., Reifenscheid, M., Blom, A.G., Wenz, A., Rettig, T., Lehrer, R., Krieger, U., Juhl, S., Friedel, S., Fikel, M., Cornesse, C. (2020), Die Mannheimer Corona-Studie: Schwerpunktbericht zur Erwerbstätigkeit in Deutschland (period covered: 20.3.-15.4.2020), Mannheim Corona Survey, Focus report: Employment in Germany. Link: https://www.uni-mannheim.de/en/gip/corona-study/. Moretti, E. (2012). The new geography of jobs. Houghton Mifflin Harcourt. Teleworkability and the COVID-19 crisis: a new digital divide? 62 Murphy, H. (2020), “Facebook to shift permanently to a more remote workforce”, Financial Times 21/5/20. Accessed at https://www.ft.com/content/4ae2fdf7-f427-4e98-abd7-feeb6b6f43f2 on 11/6/20. Newport, C. (2020), “Why Remote Work Is So Hard—and How It Can Be Fixed”, The New Yorker. https://www.newyorker.com/culture/annals-of-inquiry/can-remote-work-be-fixed Nilles, J. M. "The Telecommunication-Transportation Tradeoff: Development of Policy." (1976). Office of National Statistics (2020a), Coronavirus and the social impacts on the countries and regions of the UK, April 2020 (statistical bulletin). Office of National Statistics (2020b), Coronavirus and homeworking in the UK labour market: 2019, London: ONS. Pouliakas, K; Branka, J (2020). EU jobs at highest risk of Covid-19 social distancing: Is the pandemic exacerbating the labour market divide? Luxembourg: Publications Office of the European Union. Cedefop working paper; No 1. http://data.europa.eu/doi/10.2801/968483 Reich, R. (2020), “Covid-19 pandemic shines a light on a new kind of class divide and its inequalities”, The Guardian (26/4/20). Link: https://www.theguardian.com/commentisfree/2020/apr/25/covid-19-pandemic-shines-a-light-on-anew-kind-of-class-divide-and-its-inequalities Sandler, R. (2020), “Half Of Facebook’s Employees May Permanently Work From Home By 2030, Zuckerberg Says”, Forbes online (21/5/2020). Link: https://www.forbes.com/sites/rachelsandler/2020/05/21/half-of-facebooks-employees-maypermanently-work-from-home-by-2030-zuckerberg-says/#15dcdc684c4a. Schoenenberg, K., Raake, A., & Koeppe, J. (2014). Why are you so slow?–Misattribution of transmission delay to attributes of the conversation partner at the far-end. International Journal of Human-Computer Studies, 72(5), 477-487. Schwind, L. and Vargas Llave, O. (2019), “Telework and its effects in Europe”, in Messenger, J (ed.) (2019). Toffler, A. (1980). The third wave/Alvin Toffler. New York: Morrow. US Federal Reserve Board (2020), Report on the Economic Well-Being of U.S. Households in 2019, Featuring Supplemental Data from April 2020. Link: https://www.federalreserve.gov/consumerscommunities/shed.htm Vargas Llave, O., Mandl, I., Weber, T. and Wilkens, M. (2020), Telework and ICT-based mobile work: Flexible working in the digital age, New forms of employment series, Publications Office of the European Union, Luxembourg https://www.eurofound.europa.eu/publications/report/2020/teleworkand-ict-based-mobile-work-flexible-working-in-the-digital-age Vargas Llave, O. and Weber, T. (2020), Regulations to address work–life balance in digital flexible working arrangements, New forms of employment series, Publications Office of the European Union, Luxembourg Teleworkability and the COVID-19 crisis: a new digital divide? 63 Annexes Constructing the technical teleworkability index20 The technical teleworkability index is based primarily on the Indagine Campionaria delle Professioni, an Italian occupation survey that follows closely the structure of the American O*Net survey. We used the detailed information on tasks in this survey to classify fine-grained occupations (over 750 5-digit occupations in the Italian professional classification) as technically teleworkable or not, based on the amount of physical interaction captured by six variables, covering most of the spectrum of physical tasks. For analytical simplicity, we opted to classify occupations based on a single threshold: each 5-digit occupation is classified as not teleworkable whenever the importance score reported in the survey of any of the seven physical tasks indicators exceeds a value of 40, and as teleworkable otherwise. We settled on a value of 40 on the importance scale because it divided most accurately those occupations that involve manual and physical tasks, from those that do not – as shown by the distribution of the importance scores of each of the six variable from ICP, presented in Figure 25 below. The plot shows the density distribution of the importance scores for all 5-digit occupations, divided by one-digit occupation major group. For managerial, professional, and clerical occupations, the bulk of the distribution (i.e., most 5-digit occupations) are to the left of the 40 threshold (the dashed line) for manual dexterity, finger dexterity, general physical activity, and handling and moving objects. Most craftspeople, plant and machine operators and assemblers, and a significant share of technical professionals report values higher than 40 for inspecting equipment, structures, and materials. In addition to the six physical interaction variables contained in ICP, we also drew from the European Working Conditions Survey to add an additional physical interaction variable providing information on another type of physical interaction not covered in ICP, namely the frequency of lifting or moving people. As shown in Figure 26, the values of this indicator are zero or very low for most ISCO 3-digit occupations, but there is a distinct cluster of occupations for which it is high, mostly in health and social care. As it happens, the value for Child care and teacher’s aides (ISCO 531) fell slightly below the threshold of 40. Because this occupation has clearly much in common with all the other caring professions that report physically handling people, we decided to include it among the not teleworkable. We made another ad-hoc adjustment for cashiers in retail (CP 5.1.2.4.0, part of ISCO 523), who reported values at or below 34.8 for all physical interaction variables, including 30.4 for Handling and Moving Objects. These values appeared implausible, not only in comparison with similar occupations like ticket sellers, but also with the task descriptions of that occupation, reported in a dedicated section of the ICP. We decided to override the data-driven classification of ISCO 523, and marked them as not teleworkable. Finally, we duplicated the values of ISCO 322: Nursing and Midwifery Associate Professional into ISCO 222: Nursing and Midwifery Professional, because the official CP to ISCO mapping of occupations did not distinguish between the two. All these changes are commented and reported in the code to construct the indicators. 20 The code to construct the indices is available at https://github.com/m-sostero/telework-occupations 64 Figure 26: Values of technical teleworkability variables from ICP survey The figure shows the density distribution plot for the values of physical interaction variables across 798 different CP 5digit occupations, divided by 1-digit occupation major groups. Wider sections of the density curves mean that more occupations are clustered around that value of the importance score. The dashed line represents the threshold used to classify the technical teleworkability of occupations. Teleworkability and the COVID-19 crisis: a new digital divide? Teleworkability and the COVID-19 crisis: a new digital divide? 65 Figure 27: Distribution of physical interaction variable "Lifting or moving people" from EWCS . The figure shows the distribution of the values of Lifting or moving people for 129 different ISCO 3-digit occupations, divided by 1-digit occupation major groups. The dashed line represents the threshold used to classify the technical teleworkability of occupations. 66 Values of technical teleworkability and social interaction indices The table below shows the values of the technical teleworkability and social interaction indices for ISCO 2008 3-digit occupation groups.21 ISCO08 code Occupation title Technical teleworkability Social interaction 111 Legislators and senior officials 1.00 0.68 112 Managing directors and chief executives 1.00 0.69 121 Business services and administration managers 1.00 0.61 122 Sales, marketing and development managers 1.00 0.65 131 Production managers in agriculture, forestry and fisheries 0.00 0.62 132 Manufacturing, mining, construction, and distribution managers 0.18 0.62 133 Information and communications technology service managers 1.00 0.57 134 Professional services managers 1.00 0.67 141 Hotel and restaurant managers 0.97 0.63 142 Retail and wholesale trade managers 0.07 0.67 143 Other services managers 0.89 0.61 211 Physical and earth science professionals 0.23 0.45 212 Mathematicians, actuaries and statisticians 1.00 0.59 213 Life science professionals 0.26 0.54 214 Engineering professionals (excluding electrotechnology) 0.25 0.50 215 Electrotechnology engineers 0.00 0.51 216 Architects, planners, surveyors and designers 0.38 0.36 221 Medical doctors 0.39 0.79 222 (22) Nursing and Midwifery Professionals 0.00 0.94 225 Veterinarians 0.00 0.64 226 Other health professionals 0.59 0.75 231 University and higher education teachers 0.49 0.80 21 The table is available in csv format at https://github.com/m-sostero/telework-occupations 22 Value duplicated from 322: Nursing and midwifery associate professionals, because the official CP-ISCO mapping does not distinguish between the two. Teleworkability and the COVID-19 crisis: a new digital divide? Teleworkability and the COVID-19 crisis: a new digital divide? 67 232 Vocational education teachers 1.00 0.76 233 Secondary education teachers 1.00 0.77 234 Primary school and early childhood teachers 0.61 0.78 235 Other teaching professionals 0.86 0.74 241 Finance professionals 1.00 0.46 242 Administration professionals 0.97 0.57 243 Sales, marketing and public relations professionals 1.00 0.56 251 Software and applications developers and analysts 1.00 0.46 252 Database and network professionals 1.00 0.41 261 Legal professionals 1.00 0.43 262 Librarians, archivists and curators 1.00 0.51 263 Social and religious professionals 0.98 0.67 264 Authors, journalists and linguists 1.00 0.43 265 Creative and performing artists 0.34 0.54 311 Physical and engineering science technicians 0.01 0.45 312 Mining, manufacturing and construction supervisors 0.00 0.57 313 Process control technicians 0.02 0.38 314 Life science technicians and related associate professionals 0.63 0.35 315 Ship and aircraft controllers and technicians 0.09 0.60 321 Medical and pharmaceutical technicians 0.00 0.39 322 Nursing and midwifery associate professionals 0.00 0.94 324 Veterinary technicians and assistants 0.00 0.39 325 Other health associate professionals 0.35 0.66 331 Financial and mathematical associate professionals 1.00 0.41 332 Sales and purchasing agents and brokers 1.00 0.66 333 Business services agents 1.00 0.52 334 Administrative and specialised secretaries 1.00 0.53 335 Regulatory government associate professionals 0.53 0.57 341 Legal, social and religious associate professionals 1.00 0.74 342 Sports and fitness workers 0.04 0.67 Teleworkability and the COVID-19 crisis: a new digital divide? 68 343 Artistic, cultural and culinary associate professionals 0.11 0.48 351 Information and communications technology operations and user support technicians 0.93 0.43 352 Telecommunications and broadcasting technicians 0.00 0.32 411 General office clerks 1.00 0.39 412 Secretaries (general) 1.00 0.44 413 Keyboard operators 1.00 0.29 421 Tellers, money collectors and related clerks 0.93 0.50 422 Client information workers 1.00 0.48 431 Numerical clerks 1.00 0.26 432 Material-recording and transport clerks 0.40 0.42 441 Other clerical support workers 0.82 0.40 511 Travel attendants, conductors and guides 0.73 0.78 512 Cooks 0.00 0.48 513 Waiters and bartenders 0.00 0.56 514 Hairdressers, beauticians and related workers 0.00 0.58 515 Building and housekeeping supervisors 0.00 0.70 516 Other personal services workers 0.32 0.55 521 Street and market salespersons 0.00 0.84 522 Shop salespersons 0.04 0.80 523 (23) Cashiers and ticket clerks 0.10 0.51 524 Other sales workers 0.33 0.50 531 (24) Child care workers and teachers’ aides 0.00 0.75 532 Personal care workers in health services 0.00 0.54 541 Protective services workers 0.40 0.57 611 Market gardeners and crop growers 0.00 0.43 612 Animal producers 0.00 0.33 613 Mixed crop and animal producers 0.00 0.36 621 Forestry and related workers 0.00 0.46 23 Value for the main subgroup (CP 5.1.2.4.0) changed manually from teleworkable to non teleworkable. 24 Value just below the threshold, moved from teleworkable to non teleworkable. Teleworkability and the COVID-19 crisis: a new digital divide? 69 622 Fishery workers, hunters and trappers 0.00 0.41 711 Building frame and related trades workers 0.00 0.27 712 Building finishers and related trades workers 0.00 0.38 713 Painters, building structure cleaners and related trades workers 0.00 0.31 721 Sheet and structural metal workers, moulders and welders, and related workers 0.00 0.38 722 Blacksmiths, toolmakers and related trades workers 0.00 0.32 723 Machinery mechanics and repairers 0.00 0.33 731 Handicraft workers 0.00 0.36 732 Printing trades workers 0.33 0.30 741 Electrical equipment installers and repairers 0.00 0.40 742 Electronics and telecommunications installers and repairers 0.00 0.42 751 Food processing and related trades workers 0.00 0.50 752 Wood treaters, cabinet-makers and related trades workers 0.00 0.37 753 Garment and related trades workers 0.00 0.36 754 Other craft and related workers 0.00 0.36 811 Mining and mineral processing plant operators 0.00 0.27 812 Metal processing and finishing plant operators 0.00 0.33 813 Chemical and photographic products plant and machine operators 0.00 0.30 814 Rubber, plastic and paper products machine operators 0.00 0.28 815 Textile, fur and leather products machine operators 0.25 0.31 816 Food and related products machine operators 0.00 0.33 817 Wood processing and papermaking plant operators 0.00 0.35 818 Other stationary plant and machine operators 0.00 0.29 821 Assemblers 0.00 0.26 831 Locomotive engine drivers and related workers 0.00 0.18 832 Car, van and motorcycle drivers 0.00 0.52 833 Heavy truck and bus drivers 0.00 0.18 834 Mobile plant operators 0.00 0.25 Teleworkability and the COVID-19 crisis: a new digital divide? 70 835 Ships’ deck crews and related workers 0.00 0.26 911 Domestic, hotel and office cleaners and helpers 0.00 0.32 912 Vehicle, window, laundry and other hand cleaning workers 0.00 0.45 921 Agricultural, forestry and fishery labourers 0.00 0.24 931 Mining and construction labourers 0.00 0.19 932 Manufacturing labourers 0.00 0.24 933 Transport and storage labourers 0.00 0.20 941 Food preparation assistants 0.00 0.26 951 Street and related service workers 0.00 0.43 952 Street vendors (excluding food) 0.00 0.71 961 Refuse workers 0.00 0.29 962 Other elementary workers 0.28 0.36