Over the past decade, global robot density in manufacturing has nearly doubled. Matsuki’s study, published in 2026 in the Review of World Economics, documents a positive effect of robotization on average wage growth and at the 90th percentile, with no significant effect at the 10th percentile. The data show a positive association between robotization and total factor productivity growth, without establishing a definitive causal relationship.
Bargaining institutions moderate certain wage effects of robotization without solely determining the distribution of gains.
The Essentials
- Matsuki’s study (2026, Review of World Economics) documents a positive effect of robotization on total factor productivity with unequal distribution of gains across wage percentiles.
- Robotization is accompanied by a decline in the ratio of manufacturing employment to total employment at observed adoption rates, according to Matsuki.
- In highly unionized countries, the concentration effect is attenuated: institutions redistribute where markets concentrate.
- Matsuki’s study does not document median wage recovery during the years studied.
- By 2040-2050, democracies will need to govern a technology that enriches the global economy without enriching its majority.
Productivity Climbs, Median Wages Remain at a Standstill
The popular image of robotization is often one of job destruction. The reality is more subtle and, in certain respects, more troubling. Industrial robots have not emptied factories. They have restructured what happens within them, and who does it.
The International Federation of Robotics’ World Robotics 2025 report counts 4,663,698 industrial robots in operation worldwide at the end of 2024, a figure up 10% year-on-year. The automotive industry, electronics, and metallurgy account for the bulk of the robot fleet. Matsuki estimates a significant positive effect of robotization on total factor productivity, a result the study places favorably compared with observed sectoral averages.
Matsuki compares wage evolution according to their position in the distribution across countries, isolating the effect of robot adoption. The study finds a positive effect at the 90th wage percentile and indicates no significant effect at the 10th percentile. The study does not collect indicators relating to employment conditions.
This phenomenon is not new in economic literature. Acemoglu and Johnson theorized it: not all technologies are neutral from a distribution standpoint. Some increase demand for rare skills, others replace routine tasks without creating comparable new needs. According to this theory, robotization tends to replace rather than complement routine tasks.
Robots Compress Median Wages Without Destroying Jobs
When a robot is installed on an assembly line, it takes over the most repetitive tasks. Not all positions disappear, but the threat of substitution can reduce bargaining power for workers performing such tasks.
An operator whose gestures can be reproduced by a machine at declining cost loses their leverage. The employer no longer needs to grant a wage increase to retain this profile: the robot waits. The possibility of robot substitution can weigh on wage demands. This is what labor economists call the threat of replacement, and it need not materialize to have an effect.
Conversely, certain workers with rare skills benefit from increased demand. An engineer capable of programming, adjusting, and maintaining a roboticized cell becomes more valuable as the robot becomes more central. Their remuneration follows. Matsuki compares macroeconomic results and wage percentiles across countries, revealing unequal distribution of gains according to wage levels.
In the short term, European case studies often observe worker reallocation rather than immediate net job destruction, without establishing a single explanation for the absence of mass unemployment. Robotization affects incomes differently depending on their position in the wage distribution. Employment structures evolve under the effect of robotization. Europe faces a symmetric problem: it lacks technicians trained to manage the robots it seeks to deploy, which signals that human-machine complementarity remains a training challenge, not an automatic promise.
The Role of Unions: Technology Does Not Decide Alone
Matsuki’s most important contribution may be here: higher union density is associated with an attenuation of the adverse effects of robotization on wage growth in this study. Nordic countries, Germany, certain Canadian industrial sectors show different trajectories from those of the United States or the United Kingdom, where unionization has declined since the 1980s.
This result has a direct implication. If technology alone determined the distribution of gains, we would observe similar effects regardless of institutional architecture. The data do not confirm this hypothesis. Collective bargaining, sectoral agreements, and productivity-sharing mechanisms (profit-sharing, bonuses, wage indexation) redistribute some of the gains from robotization, showing that the distribution of gains depends heavily on institutions, regulation, and collective bargaining, but also on the characteristics and modalities of technology adoption.
Germany offers an instructive example. Co-determination (Mitbestimmung), which requires employee representation on the supervisory boards of large companies, is associated with different wage trajectories in highly automated sectors like automotive. Volkswagen, BMW, and Bosch have massively roboticized their production lines while maintaining collective agreements that redistributed a share of productivity gains. The result is imperfect, inequalities exist, and subcontractor workers often escape these protections, but the divergence is less brutal than elsewhere.
In contrast, in the United States, where private-sector unionization rates have fallen below 6%, robotization has accompanied marked wage polarization. Midwest states, the historical heart of American industry, concentrate both productivity gains and median income stagnation.
Small Industrial Enterprises in the Blind Spot
Discussion of robotization often focuses on large factories. Small and medium-sized industrial enterprises (SMEs), which nonetheless represent the majority of manufacturing employment in most developed economies, remain largely outside the analysis.
The situation of SMEs differs in two respects. Their access to robots is more limited: integration, programming, and maintenance costs remain high for small-batch production. Specialized integrators play a crucial missing-link role in enabling these companies to transition to automation without devoting disproportionate resources to it. When robotization arrives at an SME, there are fewer hierarchical levels and collective bargaining is rarer.
The risk, therefore, is that industrial SMEs concentrate both technological lag and the greatest exposure to precarity effects when technology finally arrives. Public policies to support robotization include measures for SMEs, but large companies capture a significant share of them. Their extension to smaller structures remains an issue. Expanding their reach to SMEs is a necessary condition for productivity gains to diffuse more broadly.
Wage Divergence and Its Effects on Democracies by 2040
IFR data does not allow for evaluating a U-shaped wage curve; Matsuki’s study examines certain non-linearities in the relationship between robotization and wages without validating or invalidating a general theoretical U-shape pattern. In economic literature of the 2010s, a hypothesis circulated: robotization would first destroy median jobs, then create new ones, as productivity growth diffused through the economy. The transition would be painful but temporary. Matsuki and IFR data allow for substantial qualification of this optimism.
In highly roboticized regions, available data suggest a labor market restructuring whose effects vary by institutional context. The fragmentation of the labor market has fragmented career paths. Over two generations, these effects accumulate: children of workers downgraded by automation access education with fewer resources and less social capital. Intergenerational mobility declines.
Labor economists are increasingly explicitly raising the question of compatibility between a technology that increases collective output while concentrating gains in a fraction of the population and long-term political cohesion. European studies link strong individual exposure to automation to increased support for the far right, without established generalization to deindustrialized regions of North America. The feeling of not having benefited from growth that aggregate figures describe as real constitutes unstable political terrain.
Several trajectories are conceivable by 2040-2050, though none can be termed certain. In a first scenario, collective bargaining institutions strengthen or reinvent themselves: productivity-sharing mechanisms are imposed by law or negotiation, of the type some Nordic countries are experimenting with through wage indexation clauses or collective profit-sharing. Robotization gains then feed into a more virtuous circle, provided that professional training keeps pace with evolving skill requirements.
In a second scenario, labor market fragmentation accelerates. Digital platforms and atypical contracts continue to grow, circumventing existing collective agreements. Robotization extends beyond manufacturing to logistics, hospitality, personal care, and reproduces the same polarization in sectors not yet exposed to it. Only capital taxation progresses; this is insufficient to offset stagnation in labor income.
The distinction between these two trajectories hinges on specific political choices: the level of union coverage, the capacity of education and training systems to anticipate change, the taxation of capital gains generated by automation, and the design of social protection systems, which in most developed countries remain conditional on stable employment—a work pattern that robotization transforms.
The signals to watch in coming years are well-known: the evolution of collective agreement coverage rates in newly roboticized countries, the scale of continuous training policies jointly financed by states and companies, and the speed at which robotization penetrates SMEs and services, where institutional protections are weakest.
Countries Already Redistributing Robotization Gains
A few ongoing experiments merit documentation, without overstating their reach.
In Finland, sectoral bargaining has for several years integrated productivity-sharing clauses that index part of wage increases to measured company gains. These clauses do not apply uniformly and SMEs are often excluded from them.
In Germany, several large groups have negotiated transition plans with their works councils that make robotic investments conditional on training commitments and no-layoff pledges. These plans are not generalizable to all sectors, and they presume a culture of social compromise that does not exist everywhere. But they show that robotization can be accompanied by a contractual framework that redistributes part of the gains.
In South Korea, the world’s leading country in robot density according to the IFR, with 1,012 robots per 10,000 manufacturing employees in 2023 according to IFR data (1,220 in 2024), debate on robot taxation has emerged at government level. The idea is simple: a robot can replace certain tasks or complement human work; part of the productivity generated should fund a redistribution mechanism. South Korea is exploring this avenue, though no generalized deployment has been implemented, but discussion is progressing, and it could constitute an underexploited lever for financing training and professional transitions.
These experiments share a common point: they rely on institutions capable of negotiating, enforcing, and monitoring. Weaker bargaining institutions are associated with greater wage inequality in certain data, without established proof here of a specific causal effect of robotization. Robotization interacts with existing institutional structures.
Industrial robotization is a reality that will deepen. Tools exist to redistribute its benefits, and some countries use them. The real adjustment variable through the 2040s will be political will to generalize them, and the capacity of democracies to maintain that will when those who capture the gains also have the means to influence the rules of the game.
Sources
- Matsuki T., 2026, Improving effects of industrial robot adoption, Review of World Economics, https://link.springer.com/article/10.1007/s10290-025-00626-z
- International Federation of Robotics, World Robotics 2025 Report, IFR, ifr.org
- Acemoglu D. & Johnson S., Power and Progress, PublicAffairs, 2023
- OECD, Automation and the Future of Work, updates 2026
