How new technologies are impacting the labour market

Ever since the first industrial revolution in the mid-1700s, people have been driven by a desire to produce more and cheaper goods. As we know, this development has continued right up until the present day, but instead of weaving machines and steam engines, we now have to concern ourselves with robots and artificial intelligence (AI).

In two separate studies, researchers from the Department of Economics and Business Economics at Aarhus BSS, Aarhus University, have shown how robots and AI have so far impacted working and wage conditions in the USA and Germany, respectively.

In the first of the two studies, Associate Professor Daisuke Adachi examined how Japanese industrial robots affected the labour market in the USA between 1990 and 2007.

“I’ve narrowed it down to Japanese robots for two reasons. Japanese robots account for a very large share of the US market. And the Japanese data sets for industrial robots contain an important variable, namely the price per unit, which is important if you want to show the whole picture of the impact of robots on the labour market,” explains Daisuke Adachi.

By tracking price trends for industrial robots designed to perform very specific tasks such as welding, the researcher is able to paint a detailed picture of where and when it makes financial sense for business owners to replace human labour with robots.

The researcher can analyse the impact on the labour market by comparing data from the Japan Robot Association (JARA), which accounts for approximately one third of global robot sales, with the US labour market data. The study has been published in the renowned Journal of Monetary Economics.

There are winners and losers when tasks are increasingly automated.

^{Associate Professor Daisuke Adachi, Department of Economics and Business Economics, Aarhus BSS}

Winners and losers

And the conclusion is pretty clear: Robots are not substituting all types of human labour in the same way. On the contrary. What the study clearly shows is that robot exposure is particularly high in mass production and heavy warehouse work, replacing a large share of the human workforce and thus affecting wage structures and working conditions in these occupations. Often with negative implications for low-income and middle-income groups.

“There are winners and losers when tasks are increasingly automated. Productivity increases, which is positive for the economy in general. Which is why we must also embrace technological development. But there are groups of workers who will either be replaced completely or suffer a drop in earnings due to the introduction of robots,” concludes Daisuke Adachi, who believes that the study shows that political action is needed if we are to avoid the risk of large groups of workers losing their livelihoods entirely or partially.

“Among other things, my study can be used to identify which parts of the labour market and which groups of workers are threatened by or at risk of being threatened by automation. It is important that the upskilling and reskilling of these groups is put on the agenda. In Denmark, unlike in the USA, we have strong trade unions that are aware of this, which is positive. But political leaders also need to focus on this dilemma,” says Daisuke Adachi. 

New technologies are not all the same. And their impacts on the labour market depend on the tasks they are designed to perform and who is to perform them.

^{Associate Professor Michael Koch, Dapartment of Economics and Business Economics, Aarhus BSS}

AI and robots

Much the same conclusion is reached by Associate Professor Michael Koch, Department of Economics and Business Economics, in another study published in the renowned journal Research Policy. At least when it comes to automation. Together with three fellow researchers, Michael Koch demonstrates that automation has a negative impact on wages because robots can often replace human labour.

In the study, researchers analysed the German Qualifications and Career Survey (BIBB-BAuA) for the years 2006, 2012 and 2018, in which approximately 20,000 employed people aged 15 and over and working at least 10 hours per week participated in a questionnaire survey.

However, Michael Koch and his colleagues have also investigated the use of artificial intelligence (AI) as a tool for a range of different tasks and looked at how this new technological development has affected the labour market. And this produces a different picture. When it comes to AI, at least for now, we are generally seeing a positive impact on wages for jobs exposed to AI.

“New technologies are not all the same. And their impacts on the labour market depend on the tasks they are designed to perform and who is to perform them. While robots are often suited to routine and physical tasks, AI is more focused on IT, service and development tasks. So, we’re talking about different jobs and different people with different skills,” explains Michael Koch.

Even though the researchers have identified tasks that are particularly exposed to AI-based tools, these tools will not necessarily come to replace human labour in the same way as robots, they note in the scientific article. However, there will still be a need for upskilling of employees, and politicians and business owners should be aware of this, they point out.

In addition, the researchers stress that their study focuses on the early stages of modern AI use, and that close monitoring will be needed in this area.

“AI and the use of AI are undergoing rapid development, and the situation may well change in the next few years. Job functions where AI comes to have a major impact may therefore see a negative impact on wages,” says Michael Koch.

Facts

We strive to comply with Universities Denmark’s principles for good research communication. For this reason, we provide the following information as a supplement to this article: 

Type of study	Study 1: Peer‑reviewed economics journal article combining (i) empirical estimation using an instrumental‑variable design based on robot price shocks and (ii) a quantitative general‑equilibrium model.  Data: Japanese robot shipment value/quantity by application (Japan Robot Association) used to construct robot unit prices; mapped to US occupations using O*NET task/application match scores; combined with US occupational wage data for 1990–2007.  Design/validity: Identification leverages exogenous variation in the cost of acquiring Japanese robots (“Japan Robot Shock, JRS”). To address remaining endogeneity (correlation between automation shocks and the JRS), the paper constructs a model‑implied optimal IV. Multiple specifications and robustness checks are used; estimates highlight heterogeneity across occupations.  Study 2: To carry out an in-depth analysis of the task content and the skill requirements within occupations, we make use of the German Qualifications and Career Surveys conducted over the years 2006, 2012 and 2018. The surveys are carried out over the phone by the German Federal Institute for Vocational Education and Training (BIBB) and the German Federal Institute for Occupational Safety and Health (BAuA). Each survey is based on around 20,000  employed people aged 15 and over with regular working hours of at least 10 hours per week. The BIBB-BAuA reports detailed information on worker and employer characteristics.  We also use existing measures on AI and robot exposure of occupations. Data on robot exposure at the 4-digit occupational level is obtained from Webb (2020), which is based on the similarity of robot patent texts and occupational task profiles in O*NET. AI exposure is obtained from the Dynamic Artificial Intelligence Occupational Exposure (DAIOE) index of Engberg et al. (2024).  In the final step of our analysis, we will provide worker level evidence on the implications of AI on wages. To do so, we employ the sample of integrated labour market biographies (SIAB), that is provided by the Institute for Employment Research (IAB). The SIAB is based on a 2% random sample of all individuals who have ever been registered in the German social security system.
External collaborators	Study 1: No Study 2: No
External funding	Study 1: Japan Society for the Promotion of Science (JSPS), Grant no. 23K12498.   Study 2: No
Conflict of interest	Study 1: No Study 2: No
Other	Study 1: Results reported in a peer‑reviewed article (Journal of Monetary Economics, vol. 152, article 103782, published June 2025). Key limitations to communicate: estimates depend on mapping robot applications to occupations (O*NET match scores) and on the interpretation of Japanese robot price variation as an exogenous cost shock; quantitative counterfactuals rely on the structure/calibration of the GE model. Study 2: No
Link to the scientific article	Study1: Elasticity of substitution between robots and workers: Theory and evidence from Japanese robot price data Study 2: Artificial intelligence tasks skills and wages_Worker-level evidence from Germany
Contact	Study 1: Associate Professor Daisuke Adachi, Department of Economics and Business Economics. Email: daisuke.adachi@econ.au.dk Study 2: Associate Professor Michael Koch, PhD, Department of Economics and Business Economics. Email: mkoch@econ.au.dk