Technology will transform the majority of professions in the coming years. Which workers will be most affected by technological change? This article analyses the socio-economic and political characteristics of workers at the greatest risk of finding themselves displaced by technology in the future, according to how many of the tasks typical of their current occupation could become automated. For this, the article uses a “risk of computerisation” measurement proposed by Frey and Osborne (2017), linking this to data on an individual level from the European Social Survey and comparing the risk of different sociodemographic and political profiles.

The results suggest that in Spain, risk of computerisation is relatively high and that occupations with intermediate qualifications, such as administrative occupations, will be very highly affected. Moreover, people at the greatest risk are those who are less active politically.

Growing concern exists regarding the impact of technological change on employment. Fear of the substitution of human workers by machines is of course nothing new, although to date there has never been any materialisation of the situation of mass “technological unemployment” feared first by the workers of the Luddite movement and by John Maynard Keynes one century later. The disappearance of jobs in sectors such as agriculture was compensated for during previous industrial revolutions by new job creation, either because the number of jobs in existing occupations grew thanks to increased wealth or because technological innovation led to new sectors emerging. In many cases, furthermore, technology did not lead to the disappearance of entire occupations but transformed the tasks of those who performed them, as in the case of the drivers of carts who moved on to drive motor vehicles. New technologies can create, transform or destroy employment and the net effect is difficult to anticipate (David H. Autor, 2015). 

Although the phenomenon has long roots in history, the risks arising from technological change are perhaps more tangible today, due both to the speed at which change is taking place and to the depth of the change (Brynjolfsson and McAfee, 2014). An accelerated pace of technological change can overwhelm workers’ capacity for adaptation. Although new jobs may be created that, on paper, could compensate for those destroyed, those workers “displaced” may find it difficult to acquire the technical skills necessary to perform new tasks. The “skill mismatch” phenomenon (Restrepo, 2015) is already appearing in many sectors.

Furthermore, although initially tasks susceptible to automation were of the more routine type, advances in fields such as robotics and artificial intelligence are enabling the automation of tasks that were previously considered exclusive to humans. These include tasks that require perception and manipulation, such as driving or cleaning houses, and also cognitive tasks, such as diagnosing diseases or writing legal documents. The changes are very wide-ranging in scope. 

Faced with this situation, it is essential to anticipate who will be most affected by technological change. The objective of this article is to analyse the socio-economic and political characteristics of the workers most threatened by the advance of technology in Spain. For this, the probability of computerisation measurement applied is that most used in the few studies existing on this subject, as proposed by Frey and Osborne (2017). Although this measurement has been criticised, it is arguably the best of those that exist and it offers important information if interpreted with care.

To produce this measurement, the authors asked experts in different areas which tasks will be able to be performed by machines within the next 20 years thanks to technological advances. In parallel, 70 occupations were coded as completely automatable or not, based on a description of the tasks conducted at the job. Later, the researchers used an automatic classifier to predict the probability that an occupation can be automated as a function of its current task composition. The resulting measurement reflects the proportion (from 0 to 1) of human work in each current occupation that could be potentially substituted by technological capital for 702 occupations (according to the highly detailed, 4-digit International Standard Classification of Occupations - ISCO). 

Several clarifications are necessary. Firstly, the type of tasks carried out in any occupation may vary between countries and industries, and this measurement has not been specifically developed for Spain. Secondly, the results must not be interpreted as the percentage of jobs that with all certainty will be destroyed by new technologies, but as the probability that an occupation may become automated in the near future based on the tasks carried out in that occupation. Once a task is automated the risk of jobs being destroyed increases, but what may alternatively happen is that the number is maintained, or even grows, either due to increased demand or because that occupation is transformed and incorporates new tasks. Finally, the fact that from a technical viewpoint automating a task is possible does not mean that it will necessarily happen, since automation may be too costly, or ethical or legal impediments may exist.

To cite an example, according to this measurement gardeners have a very high risk of computerisation of 0.87, but this does not mean that in 20 years the number of gardeners will fall to 13% of the current number. Perhaps gardeners will no longer cut lawns but will take care of flowers and for various reasons there may be greater demand for floral gardens. In addition, even though specialised robots may exist that can prune trees, these might be very expensive and people may prefer to continue employing people for this occupation. 

First of all the risk of computerisation is analysed in comparative perspective. To calculate this, a risk-of-computerisation value has been assigned, as calculated by Frey and Osborne, for each active person aged under 65 years (excluding the retired, people not seeking work, etc., but including the unemployed and students with a previous job) in the seventh round of the European Social Survey, the latest published to date. Differences between countries reflect the composition of the occupation; high values imply that a country has an important proportion of workers in occupations with a high risk of computerisation. The first graph shows the proportion of tasks that are currently carried out by workers and that could be performed by machines as technology advances in the coming 20 years for different European countries.

Taking into account that all the risks are situated between 0.48 and 0.58 (in a range between 0 and 1), we see that relatively little variation between countries exists. Around half of the tasks performed today by humans will be carried out by machines in a few years according to these projections. Spain has relatively high risk levels, while the Scandinavian countries or Switzerland are countries that have workers in occupations with a lower risk of computerisation. To a large extent this reflects the high investment in human capital and training carried out in these countries. 

In second place we examine the distribution of risk of computerisation within each of the main occupational groups using a box graph. In this type of graph, the central line of the box shows the median or middle value of the risk of computerisation in that occupational group. The edges of the box delimit the first and third quartile; i.e., dividing all the workers into four groups of equal size, the first quartile would include the 25% of those surveyed with a higher risk, the second reaches the value of the median, the third quartile goes from the median to the 75%, and the fourth includes the 25% of the distribution with the lowest risk of computerisation. The outer limits of the box graph show the maximum and minimum values, excluding the most extreme values or “outliers”, which appear as points on the graph. The more different these values are, the more the degree of computerisation varies within a type of occupation. 

We see that the occupations in management and “professionals” (a category that ranges from physicians to chemists, teachers and IT specialists) are those that have the lowest risk of the typical tasks of these professions being performed by machines. However, the risk is high for all the remaining categories. The so-called “technical” categories, which include professions such as specialised operators, technical engineers and inspectors have intermediate levels of training and include performance of some routine tasks. Over half of the tasks performed in these occupations could be performed by machines in years to come. The group at the highest risk is that of administrative staff, who will be easy to substitute with technological capital. Finally, other categories such as workers in services and sales, machinery operators or non-skilled workers present intermediate risk levels.

Graph 3 shows the average risk of computerisation according to the main socio-demographic characteristics. Again on this occasion, the information on the level of risk for each occupation is used and the average value for different population subgroups is calculated. 

First of all, we see that men and women have similar risk levels, although the value is slightly lower for women. This reflects the greater prevalence of occupations with interpersonal tasks, which are difficult to computerise, among women.

Notable is the major difference in the risk of computerisation according to educational attainment level and, to a lesser extent, by income. Specifically, nearly 70% of the tasks today performed by people with primary education could be performed by machines in a few years. These people are those with the highest probability of being affected by “technological unemployment” and those who will need training in new skills if they want to remain in the labour market. People who have a lower (ESO) or higher (bachillerato, BUP, COU) secondary education level also have higher than average risk-of-computerisation values. An important finding is that people who have taken Vocational Training have a lower risk-of-computerisation value than those who have taken unspecialised primary or secondary education. This seems to indicate that Vocational Training courses transmit specific skills that are valuable and distinctive. 

An important correlation also exists between risk and salary. To classify the population according to income level, quintiles are calculated in a similar way to the quartiles discussed previously: if we order the entire population from lower to higher incomes and divide it into five groups of equal size, each quintile includes 20% of the population in ascending order of income. In this case we see that the relationship between income and risk is totally progressive, so that risk declines as income increases.

Since the jobs typical of the digital economy and those of a more cognitive nature, which in principle are more difficult to computerise than routine jobs, are concentrated in the big cities, it could be expected that the risk of computerisation would be lower in the big cities. Effectively, workers in big cities present a lower risk than those in the rural environment, although the correlation with the type of municipality of residence is relatively low.

Finally, risk varies a great deal according to employment situation. Employed people present a risk of computerisation slightly below the average, whereas students and the unemployed have much higher values. To assign risk values to these last two groups, their occupation in their last job was used. We observe that the unemployed were previously in occupations with a very high risk of computerisation. This suggests that some of those currently unemployed could precisely have lost their jobs by being substituted by technological capital. 

A question very rarely explored to date refers to the political profiles of those workers at greater risk of computerisation. This question is relevant for several reasons. Firstly, it is possible that people at greater risk have political opinions different to those that have a lower risk, and, in particular, that they are more in favour of redistributing wealth, generating new demands in compensation. It is also important to observe whether those citizens directly affected participate or not in politics, as it is not very likely that their needs will be taken into account if they do not. 

Graph 4 examines whether workers with a higher and lower risk of computerisation differ in their political opinions and involvement. 

A clear relationship exists between interest in politics and risk of computerisation: people not at all interested in politics have a much higher risk. As for ideology, there is a notable difference between right-leaning citizens, with a lower risk of computerisation, and the rest. Also, workers who state that they are more against income redistribution to reduce income differences between social groups present significantly lower risks of computerisation than citizens who show themselves to be in favour of income redistribution.

As for political participation, we see that citizens who do not vote have a higher risk than those who do vote. The last variable measures how many different types of political participation activities (apart from voting) they have been involved in over the last year. Such activities may include, for example, signing petitions, collaborating with political parties or associations or donating money to political causes. We see that citizens that have not participated in any political activity except voting have higher risk-of-computerisation values than the rest. The more political activities they carry out, the more the risk is reduced. This suggests that the viewpoints of citizens at a lower risk of computerisation are those that reach the political system to a larger extent.

This article has used the risk-of-computerisation measurement proposed by Frey and Osborne (2017) together with data from the European Social Survey to analyse the socio-economic and political characteristics of workers at greater risk of computerisation in Spain. The results indicate that Spanish workers have slightly higher risk values than their peers in other countries and that this above all affects those workers with lower levels of educational attainment and the unemployed. Politically, the people at the greatest risk are those those who are favourable to redistribution and who are are more politically demobilised than other citizens.

The fact that people with a greater risk of computerisation participate notably less in politics is particularly relevant. In general, if a social group does not participate in politics, the political parties have fewer incentives to propose measures aimed at resolving the problems typical of this group. In this case, the non-participation of workers with a greater risk limits their capacity to transmit their concerns and demands to the political system. As a result, both the knowledge of political actors regarding this problem and the pressure to act may be less than if the affected workers participated to a greater extent. Whether good policies in response will actually arrive will depend on the political mobilisation of this collective, which for the time being appears to be minor.

This is a preliminary study as very little data exists on the subject at hand. In particular, our knowledge is very limited regarding basic questions such as perceptions of the risk of computerisation among the population, opinions on technology and also support for policies designed to compensate workers who may find themselves displaced, such as universal basic income for citizens. A priority for research in the future would be generating richer survey data with which to be able to explore these questions.

Autor, D.H. (2015): «Why are there still so many jobs? The history and future of workplace automation», The Journal of Economic Perspectives, 29(3).

Brynjolfsson, E., y A. McAfee (2014): The second machine age: work, progress, and prosperity in a time of brilliant technologies, Nueva York: W.W. Norton.

European Social Survey Round 7 (2014), Edition 2.1

Frey, C.B., y M.A. Osborne (2017): «The future of employment: how susceptible are jobs to computerisation?», Technological Forecasting and Social Change, 114.

Restrepo, P. (2015): «Skill mismatch and structural unemployment, », Massachusetts Institute of Technology.