The automation probability is a percentage value and equals the proportion of essential tasks (activities) in an occupation which, in theory, could already be performed by computers or computer-controlled machines. In this regard, research speaks about the substitution potential of the occupation. The Federal Employment Agency asked occupational experts to determine which tasks are essential for an occupation on the basis of training regulations or job vacancies.
On the basis of the technological possibilities available in 2013, Dengler and Matthes (2015) determined for the first time for each task included in BERUFENET whether they could potentially be performed completely automatically by computers or computer-controlled machines. Because many new technologies have been developed since then and tasks that were previously considered non-replaceable could potentially be performed by robots or computer programmes today, a reassessment for the technological possibilities in 2016 was made and the substitution potential was updated. The update, however, did not just consider the changes in the technological possibilities but also that the job profiles in some occupations have changed and that new tasks or occupations have come about (Dengler/Matthes 2018). The determination of the substitution potential is exclusively geared towards the assessment of the technical feasibility. The substitution potential of an occupation results when you divide the number of tasks of an occupation which can be automated by the total of all its tasks and multiply this by 100.
The Job-Futuromat takes into account only the essential tasks usually to be performed in the selected occupation. In addition to these, other tasks can also be important in a job. Because they also affect the substitution potential, the result of the Job-Futuromat should not be misunderstood as an indication of how likely it is to be replaced by a computer or computer-controlled machines in a job.
No, the Job-Futuromat does not make a forecast. The substitution potential in the Job-Futuromat does not - as other online tools or studies - specify the probability that an occupation will disappear in the next 10 or 20 years because it will be performed by computers or computer-controlled machines. It rather indicates which percentage of the tasks in the occupation can already be performed by computers or computer-controlled machines in theory.
If a task is replaceable in the Job-Futuromat, this, however, does not mean that it is already performed by computers or computer-controlled machines. Perhaps human work is more economic, more flexible or of better quality. For example, almost every shop employs cashiers although the substitution potential of the occupation is 100 %.
The substitution potential therefore only indicates the potential that parts of an occupation could be replaced by computers or computer-controlled machines in theory. In doing so, the Job-Futuromat, however, says quite a bit about the future of occupations. It is very likely that these occupations and the tasks performed in these occupations will change. While the replaceable tasks will be performed by computers or computer-controlled machines, the non-replaceable tasks will still have to be performed by humans; and new tasks will emerge: the new machines must be operated, monitored, maintained, built and (further) developed. The change in the occupations means that further training will become more important.
The substitution potentials are snapshots. For the first time, Dengler and Matthes (2015) determined the replaceability of individual occupations on the basis of technological possibilities in 2013. Because a number of new technologies have now reached market maturity (especially mobile, collaborative robots, self-learning computer programmes as well as the first applications of 3D printing and virtual reality), a reassessment became necessary. The currently reported substitution potentials in the Job-Futuromat relate to the technological possibilities existing in 2016 ( Dengler/Matthes 2018 ). However, it is foreseeable that, also in the future, other currently non-replaceable tasks will become replaceable tasks because technological development progresses.
In the last few years, however, not only the technological possibilities have changed but also the task profiles of the occupations. In some occupations, tasks are no longer relevant since they are performed by robots or computer algorithms. The introduction of new technologies, however, also leads to the development of new tasks that become indispensable for the exercise of an occupation such as mastering new software. In addition, accelerated technological change leads to the emergence of new occupations.
However, this does not necessarily mean that a task that has been classified as replaceable for the moment will actually be replaced in the next few years. Perhaps human work is more economic, more flexible or of better quality in this case. Also, legal or ethical obstacles may oppose automation.
No. However, the job could change.
There are occupations with a substitution potential of 100 %, e.g., the occupation of the cashier, in which employment has grown in recent years nonetheless. In fact, the number of employees subject to social insurance contributions in the type of occupation “cashiers and ticket agents” increased in 2012 and 2015 even by over 8 % - and thus more than the trend of all occupations (+ 5.5 %).
There are also occupations with a substitution potential of 100 % whose employment numbers are dwindling. However, if you take a look at all the occupations whose substitution potential is considered “high” (more than 70 %), then you will see: these occupations, too, show an increase in employment - albeit to a lesser extent than occupations with a lower level of substitution potential.
The reason for this: Tasks which could already be performed by computers or computer-controlled machines are often still being performed by humans for some time - or even permanently. Because as long as human labour is more economic, more flexible or of better quality, or legal or ethical obstacles stand in the way of automation, there is a general tendency against automation.
Even if tasks are actually replaced, this does not necessarily mean that an occupation disappears because of that. However, it is likely that the tasks performed in this occupation will change: While the computers or computer-controlled machines perform replaceable tasks, the non-replaceable tasks will gain importance for the humans, and new tasks will emerge. Continuous further training is, therefore, gaining importance.
Even in occupations with a low substitution potential - even 0 % - in the future, it will be quite possible that tasks will be replaceable which we still cannot imagine that they will ever be performed by computers or computer-controlled machines.
The extent to which the ongoing digital transformation will reduce jobs in the long term is a controversial and open question in research. Dengler and Matthes have calculated that the proportion of employees subject to social insurance contributions working in occupations with a high substitution potential - i.e., in occupations in which at least 70 per cent of tasks could be performed by computers or computer-controlled machines - increased from 15 per cent in 2013 to 25 per cent in 2016.
This increase, however, does not necessarily mean that a lot of people will lose their jobs in the future by the ongoing digital transformation. These calculations do not consider that digital transformation leads to new products and services being offered on the market or that innovations and productivity growth lead to price reductions that could lead to an overall increase in employment instead of a reduction in employment. It is also impossible to forecast how much the task profiles of the occupations will change and which new tasks and occupations will emerge - adaptation mechanisms which are extremely relevant for managing the technological change, because occupations can change in such a way that they can keep pace with the digital transformation more or less well.
It is estimated that the acceleration of the technological change, on the one hand, will lead to 1.5 million jobs being cut, but, on the other hand, just as many jobs will be created (Wolter et al. 2016). Therefore, the challenge of the digital transformation is, above all, to assist the changes in the occupational and requirements structure associated with the introduction of new technologies. Knowledge becomes obsolete more and more quickly so that initial training will suffice more and more rarely to cope with the requirements of one’s entire working life. Learning must become the norm in working life.
For the nearly 4,000 occupations available in the Job-Futuromat, the Federal Employment Agency asked occupational experts to determine which tasks are essential for an occupation on the basis of training regulations or job vacancies. These tasks and more information are available online for all occupations in the expert database BERUFENET of the Federal Employment Agency.
The Job-Futuromat takes into account only the essential tasks usually to be performed in the selected occupation. In addition to these, other tasks can also be important in a job. Because they also affect the substitution potential, the result of the Job-Futuromat should not be misunderstood as a forecast of how likely it is to be replaced by computers or computer-controlled machines in a job.
Dengler and Matthes ( 2015, 2018 ) have determined for each task whether they could already be performed by computers or computer-controlled machines following programmable rules. They considered only the technical feasibility. In that, three encoders have independently researched whether there is a computer or a computer-controlled machine which could currently perform a concrete task from the total of approximately 8,000 tasks included in BERUFENET fully automatically. For instance, driving buses, cars or trucks - in contrast to rail-bound vehicle operation - is still considered as non-substitutable because the corresponding technologies are still being tested and therefore one cannot yet speak of market maturity.
When you run the Job-Futuromat for an occupation for the first time, the substitution potential for the selected occupation will be calculated. Each task has the same influence in this - it will be weighted equally. This corresponds to the initial central position of all sliders.
Since not all tasks are performed with the same frequency in actual daily routines, you can use the sliders to adjust their frequency:
Moving the slider to the right means that this task is performed more frequently; moving the slider to the left means that the task is performed less frequently. The substitution potential increases when a replaceable task is performed more frequently, and decreases when it is performed less frequently. On the other hand, the substitution potential decreases when a non-replaceable task is performed more frequently, and increases when it is performed less frequently.
If the substitution potential in the selected occupation is 0 %, the sliders have no influence: this means that all tasks in an occupation are non-automatable. Also re-weighting the tasks does not change the substitution potential then.
If the substitution potential in the selected occupation is 100 %, moving the sliders also does not change anything, because even re-weighting the tasks does not change the fact that all tasks are replaceable in theory.
What is important for the recalculation of the substitution potential is how the sliders are compared to each other. The further to the right a slider is as compared to the other sliders, the higher is the weighting with which the respective task increases or decreases the substitution potential. The substitution potential is mathematically calculated as follows: If a task is never exercised, it receives the value 0; if it exercised constantly, it receives the value 1. For the slider position in between, it receives a value between 0 and 1. The substitution potential is the sum of all these task values, whereby each of them is multiplied with 1 when a task is substitutable, otherwise with 0. Before this sum is calculated, all task values are standardised in such a way that their total is 1 (i.e., all divided by the same number). Hence, the individual task values become “weights” which result in a substitution potential between 0 and 1 (or between 0 % and 100 %).
The Job-Futuromat contains information for approx. 4000 individual occupations. These are occupations known in Germany according to the BERUFENET expert database of the Federal Employment Agency.
Statistics and research do not refer to these individual occupations just as “occupations” but as “core occupations”, which again include more than 20,000 occupational titles. These are summarised in the so-called Classification of Occupations 2010 (KldB 2010) into types of occupations, occupational groups and other superordinate categories. In this, occupational titles which are similar regarding the skills and knowledge relevant for exercising the occupation are grouped together.
The number of employees includes both full- and part-time employment (including trainees). Only employees subject to social security contributions are considered. For occupations with a high proportion of self-employed persons, civil servants or people in marginal employment, the actual number of people in this occupation might be much higher.
The data come from Federal Employment Agency statistics and are based on employer notifications. They apply for 31 December of each year respectively.
The data do not relate to individual occupations but to types of occupations. The almost 4,000 occupations which can be found via the search field of the Job-Futuromat are grouped together into approximately 1,200 types of occupations. Federal Employment Agency statistics only has the numbers of employees for these types of occupations available, not for individual occupations.
If there were less than three employees in an occupation in a year, no data will be displayed for reasons of data protection.
The mean wage describes the mean monthly gross wage, represented as so-called median, i.e., half of the employees receive a smaller wage, the other half of the employees receive a higher wage. It is calculated for all full-time employees subject to social security contributions in any type of occupation (excluding trainees). It is based on the annual gross wage which employers report to Federal Employment Agency statistic until 31 December of each year (see Methods report). For occupations with a high proportion of self-employed persons or civil servants, the values can deviate from that.
Individual wages can differ greatly from the median sometimes. This is why the fluctuation range around the median, within which 50 % of all wages lie, is specified in addition. The lower end corresponds to the first quartile (the limit between the first and second quarter of all employees - sorted by the monthly gross wage); the upper end corresponds to the third quartile (the limit between the third and fourth quarters of all employees - sorted by the monthly gross wage).
If the monthly gross wage of a full-time employee exceeds the so-called “contribution assessment ceiling”, Federal Employment Agency statistics will not include this exactly. If the wage is above this threshold for more than half of all employees, the median cannot be calculated. In these cases, only the contribution assessment ceiling will be specified. It changes annually and was € 6,200 per month in West Germany and € 5,400 per month in East Germany in 2016. The representation is based on the lowest contribution assessment ceiling applicable to the federal territory for East Germany (>€ 5,400).
The data do not relate to individual occupations but to so-called types of occupations. The almost 4,000 occupations which can be found via the search field of the Job-Futuromat are grouped together into approximately 1,200 types of occupations. Federal Employment Agency statistics only has the monthly gross wages for these types of occupations available, not for individual occupations.
If, in one year, there was wage information for less than 1,000 employees in one type of occupation, no wage will be displayed since Federal Employment Agency statistics classifies their calculation as not reliable.
Currently, the Job-Futuromat reports the occupational substitution potentials for the technological possibilities in 2016 (Dengler/Matthes 2018). The Job-Futuromat is based on the occupational information of approximately 4,000 individual occupations. These come from the BERUFENET expert database of the Federal Employment Agency. The information on the numbers of employees and monthly gross wages come from Federal Employment Agency statistics.
The Job-Futuromat was developed in the context of the “Zukunft der Arbeit” (the future of work) ARD theme week in 2016 in a cooperation of the ARD, the Institute for Employment Research (IAB) and the Federal Employment Agency. The Federal Employment Agency statistics and the IAB provided the data for the tool and professionally advised on the design of the Job-Futuromat. The concept of the tool originally came from data journalist Björn Schwentker. The “Thadeus Roth” agency is responsible for programming and design.
Since November 2017, the Job-Futuromat has been continued in a modified design under the responsibility of the IAB. The original results referred to the technological possibilities in 2013. Because many new technologies have been developed since then and tasks that were previously considered non-replaceable could potentially be performed by robots or computer programmes today, a reassessment for the technological possibilities in 2016 was made. That is why the updated substitution potentials for the technological possibilities available in 2016 can be retrieved in the Job-Futuromat since February 2018.