The invention of the transistor in 1947 ignited the spark for the next step in humanity’s economic transformation. The transistor served as the building block for the microchip, which in turn enabled the development of digital technologies. The first commercially available computer – a simple machine – was designed to replace punched-card accounting machines that were used to input data in the form of punched holes. Now, 75 years later, computer technology has become increasingly ubiquitous and is shaping most aspects of contemporary living. Digital technology has transformed the way governments, companies and individuals interact and work. However, with the ongoing processes of digital transformation, many are wondering how the adoption of new technologies is going to affect the labour market. This has led to controversial debates about the potential of digitalisation to reshape work, including the creation or destruction of jobs. Ultimately, the increasing adoption of artificial intelligence (AI), robotics and digital technology in general has prompted fears that digitalisation may yet ignite a new step of economic development, in line with Rifkin’s thesis on the end of work. 

Over the last few decades, concerns about the risks of computerisation and technologically driven unemployment have received academic attention through various studies. These studies often argue that technological change is biased towards certain skills, which can polarise wages and exacerbate existing inequalities. Most significantly, this can have an impact on medium- and low-skilled workers, increasing their risk of being displaced. However, the relationship between digitalisation and the displacement of jobs is not homogeneous across all occupations. Consequently, understanding these processes requires breaking down the elements that constitute occupations, allowing for a more granular perspective of how digitalisation evolves in this context. Therefore, rather than focusing on occupations, the study adopts a task-based approach, which views an occupation as a bundle of tasks. This focus on tasks, in addition to occupations, reflects the distinct technical nature of the production process, and thus is better suited for objectively assessing the impact of digitalisation on jobs.

Naturally, some tasks may be more vulnerable to technological change than others. Computer technology is known to be particularly suited to execute algorithms, i.e., performing a set of sequential tasks in a pre-determined order that follows specific rules. Thus, it can be assumed that routine tasks should be easier to automate since they can be broken down into sequential activities and programmed accordingly. By breaking down occupations into the tasks that comprise them and focusing on the relationship between digitalisation and the routinisation of tasks, it is possible to obtain a clearer understanding of the impacts of digitalisation on work regarding automation. The study on which this article is based aims to bring greater visibility to these processes, in the Portuguese context. 

Building on the above discussion, the study seeks to answer two questions:

• Is there a disproportionate distribution of skilled to unskilled workers between more and less digitalised occupations in Portugal? 

• Are low-skilled workers more at risk of digital automation, based on the level of digitalisation and the number of routine tasks that make up each occupation?

The researchers address these questions through an innovative approach, based on the work by Cirillo et al. (2020), and implement two indices that seek to characterise occupations according to their level of digitalisation – Digitalisation Index (DI) – and routineness – Routineness Index – (RI).

Occupations vary substantially in terms of work content and the corresponding tasks. The technical and organisational characteristics of work, often specific to each job or occupational category, contribute to this heterogeneity. Therefore, it is not easy to assess the level of digitalisation of occupations. To address this challenge, the authors developed a Digitalisation Index (DI), which evaluates the proportion of digital tasks that make up occupations and the importance of digital technologies in their completion. By applying this index, they found significant differences in the use and importance of digital technology across occupations in Portugal. For instance, figure 1 shows that higher-skilled occupations, such as managers, professionals or technicians, are characterised by the highest levels of digitalisation (DI > 0.5). In contrast, medium- and low-skilled occupations, except clerical support workers, are characterised by considerably lower levels of digitalisation (DI < 0.5). Thus, the researchers found evidence of skill-biased digitalisation in Portugal, i.e., digitalisation is more prominent across occupations characterised by high-skilled workers.

In order to address the risk of automation of occupations linked to the routineness of occupations, the study implemented a Routineness Index (RI) (Autor et al., 2003) which focuses on cognitive and non-cognitive tasks to determine the corresponding level of routinisation. Through the implementation of this index, the researchers found no relationship between skills and the routineness of tasks that make up occupations in Portugal. In fact, looking at figure 2, it can be seen that most occupations have a low Routineness Index (RI < 0.5), except technicians, clerical support workers and plant and machine operators. This lack of relationship between skills and routineness level suggests that automation in Portugal is not skill-biased.

The authors assessed the risk of digital automation of occupations in Portugal by analysing the relationship between digitalisation and routinisation. The results suggest that occupations with high levels of digitalisation and routinisation have a higher risk of digital automation. Figure 3 illustrates this relationship. Each dot represents a specific sub-occupation traditionally grouped into occupations (managers, professionals, technicians, etc.). The occupations considered most at risk of automation are those with a higher concentration of sub-occupations in the upper right quadrant. Although routinisation is apparently heterogeneous, some interesting trends emerge when it is combined with digitalisation. Overall, most occupations have a low exposure to the risk of digital automation either because of the low level of digitalisation of tasks (DI < 0.5) or because they are not very routinised (RI < 0.5). However, two occupational categories – clerical support workers and technicians – stand out from the rest as having a higher risk of digital automation (DI & RI > 0.5). Notably, these occupations are often highly dependent on digital technologies and characterised by a certain degree of routineness, even though they are not low-skilled.

Mainstream economic theory highlights the role of technological change in cementing wage differences between workers. Traditionally, high-skilled workers benefit the most from the introduction of new technologies, which are frequently seen as complementary to their work. Conversely, low-skilled workers are at a higher risk of being replaced by the same technologies. This phenomenon is then translated into an increasing demand for high-skilled workers, hence the term skill-biased technological change. In this model, technology is assumed to have a homogeneous impact on the labour market structure, blindly increasing the demand for high-skilled workers.
Following the skill-biased approach, a new and more refined perspective was proposed to explain the impact of technological change on the labour market, with a particular focus on computerisation. This new perspective focuses primarily on how technology determines the extent to which production tasks are allocated to labour capital. Essentially, proponents of this approach argue that digitalisation reallocates a particular set of tasks, often characterised by high levels of routine, from labour to digital technology capital, hence the term routine-biased technological change. In this regard, the level of routinisation of work supersedes skills in assessing the potential for digital automation.

Both of the approaches described above resonate with the results of the research on which this article is based. The data shows evidence of skill-biased digitalisation in the Portuguese workforce. The relationship between occupational skills and their level of digitalisation is clearly positive, i.e., high-skilled occupations are more digitalised. However, the researchers also found that occupations characterised by low-skilled workers are not necessarily associated with a higher risk of digital automation. This is important as digital automation has been a primary concern expressed by researchers and policymakers when looking at the risk of digitalisation destroying jobs. However, a first reading of the data suggests that most occupations in Portugal are still characterised by a low risk of digital automation, and therefore a low risk of job substitution. Nevertheless, this might be a signal that the tasks are still characterised by a low level of digitalisation or routinisation. Furthermore, and perhaps more surprisingly, the occupations with the highest risk of digital automation are, in fact, medium- and high-skilled occupations: clerical support workers and technicians, respectively.

In 2021, technicians and clerical support workers represented approximately 21.9% of employees in Portugal. Thus, a high risk of digital automation for these occupational categories is still a concern. However, it is important to bear in mind that these processes will not happen overnight. Moreover, digitalisation is not a uniquely destructive process. Naturally, digitalisation may lead to the automation of some tasks, rendering particular bundles of tasks obsolete. However, it is also expected that it will create a demand for new tasks and occupations. Thus, a comprehensive assessment of the potential impact of digitalisation goes beyond a technical reading of occupations and tasks and requires looking into the specific context of workers who will be affected by these transformations.
In this respect, the data from the survey shows that although the overwhelming majority of respondents are optimistic about the future of their jobs, low-skilled workers are more concerned about what an increasingly technology-mediated working environment could bring to their future. This is not surprising, given that low-skilled workers are less prepared to adapt to the changing conditions that naturally emerge as the digitalisation of jobs increasingly becomes the norm. Looking at the particular case of technicians and clerical support workers in Portugal, the fact that these workers have a higher level of skills should also indicate that they are better equipped to deal with the transformations brought about by technological change. Given the correlation between high skills and digitalisation, it can also be inferred that because these workers’ tasks are highly mediated by technology, they are more accustomed to, and therefore less pessimistic about, the changes that technology can bring to their working lives. Conversely, as low-skilled workers interact less with technology, they are more anxious and suspicious of technological change. 

To conclude, some important considerations should be taken into account for policy development. Although digital automation has accelerated in the last few years, it is unlikely that its impacts on job destruction will occur suddenly. Occupations are made up of different bundles of tasks that may or may not be automated by technological advancements. Job substitution does not depend on the level of routine and automation of a particular task but rather of a set of tasks, and hence is not the obvious result of automation processes. As tasks are replaced by technology, new tasks may also be assigned to workers, possibly including the role of maintaining the new technological infrastructure. These processes suggest that as technological change increases, workers, organisations, and policies in general will need to adapt rapidly to keep up with these developments. While public policies should promote conditions for the technological upgrading of organisations, and their digital transformation, they must also support complementary innovation in organisational processes and the adaptation and training of human capital to adjust to wider changes in technology and the corresponding skills base. These policies should aim to mitigate the negative impacts of technological advancements on work and employment while promoting the benefits that arise from this process. Investment in human capital is key to equipping workers with the necessary tools to face technological progress. New skilling, reskilling and upskilling will play a crucial role in this endeavour. 

The analysis and results presented in the study are a direct consequence of the methodological approach used. In this respect, the choices made by the authors have influenced the results and their interpretation. There are three choices of particular relevance. First, they decided to use a task-based approach. The primary reason for this choice was that it disaggregates occupations into a more realistic and suitable unit of analysis for measuring and quantifying the impact of technological change on labour. However, occupations are more than tasks. This approach does not take into account other important aspects of the work context, such as sociability, which is an intrinsic part of workers’ daily lives. Second, their analysis focuses primarily on digital automation rather than robotisation, as the former is particularly characteristic of the more recent wave of technological change. Considering that robotisation acts mainly as a substitute for low-skilled workers and that, according to their survey, the routinisation of low-skilled jobs is not negligible, the corresponding analysis could provide additional perspectives on this issue. Finally, their data was collected mainly at the ISCO-08 2-digit occupational category level. This smooths out any sharp differences that may arise when considering occupations at a more granular level.

AUTOR, D. H., F. LEVY, e R. J. MURNANE (2003): «The Skill Content of Recent Technological Change: An Empirical Exploration», The Quarterly Journal of Economics, 118(4), 1279-1333. 

CIRILLO, V., et al. (2020): «Digitalisation, routineness and employment: An exploration on Italian task-based data», Research Policy, 50(7). 104079.