By Glenda Quintini and Guillermo Montt.
A range of OECD analysis (see the recent OECD Policy Brief on The Future of Work) has been exploring the relationship between digitalisation, jobs and skills, the magnitude of potential job substitution due to technological change, the relationship between globalisation and wage polarisation, as well as the changes to the organisation of work.
Technological change and digitalisation often raise fears that workers will be replaced by computers and computer-enabled robots, resulting in what has been called technological unemployment. These fears have not materialised for past technological advances as the creation of new jobs outweighed the labour-saving impact of technology. However, it has been argued that recent and future advances in computing power and artificial intelligence may lead to the automation of a much broader range of tasks than just routine tasks, including those that were previously the exclusive domain of humans, such as reasoning, sensing and deciding. This could result in a much more profound and disruptive impact on employment than during previous episodes of major technological innovation. Frey and Osborne (2013) have estimated that 47% of US workers are in occupations that could be performed by computers and algorithms within the next 10 to 20 years. Using the same methodology, similar results have been found for European countries, ranging from 35% of jobs at risk of automation in Finland to 59% in Germany, with the variation across countries reflecting differences in occupational structures.
However, these studies disregard the considerable differences in the tasks that are performed across jobs within the same occupational title. These differences are accounted for in the study by Arntz, Gregory and Zierhan which was commissioned by the OECD. As reported in the recent OECD Policy Brief on The Future of Work, this study exploits data from the Survey of Adult Skills (PIAAC). Rather than assuming that occupations are displaced by machines, the authors argue that certain tasks can be displaced. To the extent that bundles of tasks differ across workers within occupations, the proportion of jobs at risk of automation may be somewhat lower than estimated by Frey and Osborne who assume that all workers in an occupation face the same risk of being displaced from their jobs because of digitalisation.
These estimates follow a task-based approach to transfer the results by Frey and Osborne to other OECD countries. More specifically, they estimate the relationship between workplace tasks in the US and the risk of automation that Frey and Osborne determined by consulting experts. They then use this statistical relationship to transfer the potential for task automation to other OECD countries. These estimates rely on individual data from the Survey of Adult Skills regarding a comprehensive list of tasks that people actually perform at their workplace. Using individual-level data, they take account of the fact that individuals within the same occupation often perform quite different tasks. Moreover, the task structures are self-reported by the individuals and thus likely a better indicator of workers’ actual tasks than occupational descriptions.
When using this task-based information, the share of jobs at a high risk of automation is just 9% in the United States and ranges between 6 and 12% in other OECD countries participating in the Survey of Adult Skills (Figure 1). In each job, the frequency of tasks relating to the use of reading, writing and complex ICT at work significantly reduces the likelihood of automation. On the other hand, tasks involving physical dexterity increase the likelihood of automation. A significantly larger share of jobs, ranging from 20-30%, faces a medium risk of automation. These are jobs where many – but not all – tasks are at risks of being automated. The jobs may not be replaced entirely, but could be significantly retooled.
Overall, the tasked-based approach is less restrictive than occupation-based approaches, which rely not only on the assumption that task structures are identical for all jobs with the same occupational title but also on the assumption that occupation task structures are the same in the US and other countries. The procedure still assumes, however, that workers with the same task structure face the same automation risk in all OECD countries and regions participating in PIAAC.
Differences in the risk of automation of jobs across OECD countries and regions are only partly due to differences in the industry or occupation structure of their workforce. Since workers in the same industries or occupations tend to perform different tasks in different countries, they can be more or less at risk of seeing their job automated. This is particularly the case with tasks related to work organisation – such as instructing, training and teaching others, planning one’s own activities, influencing others – which are found to be associated with differences in the risk of automation across countries. Some of these country differences may also be explained by differences in the extent to which countries have already invested in new automation technologies – i.e. the extent to which countries have already replaced labour by capital for performing certain job tasks.
One caveat to exercises of this kind – stressed in both the OECD Policy Brief and by Frey and Osborne – is that readers should not confound the potential for automation with actual employment losses. In particular, the technical possibility to use machines rather than humans to carry out certain tasks need not mean that the substitution of humans by machines actually takes place. In many cases, there are legal as well as ethical obstacles that may prevent such a substitution or at least substantially slow down its pace. Moreover, the substitution may not be reasonable from an economic point of view. However, even in the absence of such obstacles, workers may adjust to a new division of labour between machines and humans by switching tasks.