Industrial Robots in Europe: Higher Density Does Not Prove More Jobs or Productivity

European industrial automation is often summed up in a single figure: the number of robots installed for every 10,000 people employed in manufacturing. It is a useful indicator for comparing the intensity of automation, but it does not measure how many jobs are lost or created, or how much output per hour increases. Confusing those outcomes turns a descriptive statistic into a causal conclusion that the figure does not contain.

The IFR reported that in 2024 robot density reached 231 units per 10,000 manufacturing employees in the EU-27. For Western Europe it reported 267, compared with 204 in North America and 131 in Asia; the global average was 132. The denominator is manufacturing employment, not the entire working population, and regional comparisons are not equivalent to a ranking of productivity or workers’ wellbeing. (IFR, World Robotics 2025)

The figure is not a count of cobots, either. The release refers to industrial robots in general and does not break out how many units can operate collaboratively with people. The data therefore provide context for factory robotisation, but cannot show that collaborative robots are becoming more common at the same rate or estimate their specific effects on employment.

The World Robotics 2025 report contains density data for 2024. The IFR defines the measure as operational robots in relation to the number of people employed in manufacturing. According to its release, Western Europe’s density rose 3% year on year, while the EU-27 remained above the global average. These are relative stock figures, not a series of hires, redundancies or wages attributable to robots.

Geography and timing also need to be kept separate. “Western Europe” and the “EU-27” are not identical groups; moreover, figures for 2024 published in a 2025 report should not be presented as measurements of the 2025 labour market. The IFR offers a comparative snapshot of automation; tracking employment trends would require labour data matched by country, sector and period. (IFR)

Density may rise because more robots are installed, because the number of manufacturing employees changes, or because both happen. That makes it useful for understanding the scale of technological adoption, but also limits what can be inferred from it alone: a ratio does not identify the mechanism behind a change. It does not show whether systems replaced tasks, complemented existing teams or accompanied an expansion in production.

A collaborative robot is a type of industrial robot designed for particular forms of interaction with people. The label does not mean that any shared task is automatically safe: assessment depends on the application, workspace, attached tool and work organisation. Aggregate industrial-robot density therefore cannot tell us how many collaborative workplaces exist or what risks they involve.

The distinction matters when analysing work. A robot in an isolated cell and a cobot assisting at an assembly station may automate different tasks and affect work pace, physical strain and required skills in different ways. The technology category alone cannot describe workers’ experience; task, process, exposure and occupation data are needed from before and after deployment.

Business sources describing cobots highlight their use in repetitive tasks and potential gains in flexibility, but suppliers’ claims do not replace an independent assessment of outcomes. Nor do they justify generalisations about net employment across the EU. To assess a particular installation, it is worth asking which tasks change, who supervises the system, what training is offered and how safety is verified. (Bama, explanation of collaborative robotics)

To estimate productivity per hour worked, output or value added must be related to hours worked in a defined sector and period. Robot density alone supplies neither of those variables. It can inform a comparison, but it does not prove that robots caused productivity to improve: companies that adopt automation may differ from those that do not in investment, scale, demand, product mix or workforce skills.

A 2026 academic study of selected EU countries and sectors examines the relationship between robot density and value added using panel data and a quantile-regression method. Its abstract reports a positive association, especially in higher-performing sectors, and varying effects from R&D and human capital. This is not proof that every robot has the same effect, nor an estimate specific to cobots; the article studies sectoral robot density. (Doran et al., Automation and growth in the European Union)

The distinction between association and causation matters for public- and business-sector decisions. A high-value-added sector may have more resources to automate; in turn, automation may contribute to its performance. To distinguish these directions, an analysis must explain its methods, controls, coverage and limitations. Without that context, presenting a correlation as a causal effect overstates the available evidence.

Automation often affects specific tasks first, not necessarily entire occupations. A system may take over repetitive movements while a worker continues to handle adjustments, quality control, troubleshooting or coordination. It may also reorganise shifts and teams, or make production viable when it was not previously possible. The net employment effect depends on how activity changes, as well as demand, investment and workers’ transition into other roles.

The European Commission presents robotics as a technology with the potential to support productivity, competitiveness and job creation, while also linking it to social challenges and the need for safe, responsible adoption. That institutional position is not a measurement of cobots’ net effect on employment. (European Commission, Robotics)

It is therefore not sound to claim that greater robotisation always destroys jobs, or to maintain that it necessarily creates more jobs than it displaces. Opportunities and costs may be unevenly distributed across firms, regions and workers. A useful analysis should distinguish total employment, hiring and separations, wages, hours, training and changes in tasks; it should also track these indicators over time instead of inferring them from robot density.

Companies and public authorities can build a more informative assessment by combining technological, production and labour measures. The aim is not to attribute every change to a robot, but to observe what happened, in what context and to whom. A before-and-after comparison is a starting point, but it is stronger when it considers similar processes that did not adopt the system and documents concurrent changes in demand, investment and staffing.

A minimum monitoring list could include:

• Output and productivity: conforming units, value added and hours worked, using comparable periods and units.
• Employment and job quality: workforce size, hires, departures, wages, hours, absenteeism and changes in roles.
• Tasks and capabilities: activities automated, human intervention, training provided and time needed to acquire skills.
• Safety and organisation: incidents, risk assessments, ergonomics, work pace and employee participation.
• Distribution of outcomes: differences between profiles, sites and worker groups, including who bears costs and receives benefits.

The baseline should be established before the system is installed, and indicators should be tracked for long enough to detect adjustments and delayed effects. For cobot projects, it is particularly important to record the specific application: the same robot model can perform different tasks depending on its tools, programming and environment. Measuring the application prevents one case from being treated as representative of an entire technology category.

The IFR data put the EU-27 at 231 industrial robots per 10,000 manufacturing employees in 2024, above the global average. This is relevant for comparing the intensity of automation, but it does not identify how many are cobots, what tasks they perform or how they change employment or productivity. Responsible interpretation begins by respecting that scope.

Assessing social impact requires compatible labour and production statistics, sector-level analysis and evidence about the tasks actually reorganised. Training, safety and the distribution of benefits also need to be observed. The evidence available here can describe robot density and cite research on associations between robotisation and value added; it cannot establish that an increase in robots, by itself, causes higher productivity or employment.

The practical question for a company is not only how many robots it has installed, but what problem they solve, what work changes and how gains and costs are shared. For public policy, the challenge is to support skills and transitions without assuming that aggregate benefits automatically reach every worker. Density is an adoption indicator; labour-market impact must be measured separately.