South Korea has 1,066 industrial robots per 10,000 manufacturing employees, according to the International Federation of Robotics, seven times the global average of 152. The distribution of automation gains depends notably on institutions, but also on sectoral structure, productivity, companies, and the labor market.
The essentials
- South Korea is the world’s most roboticized economy with 1,066 robots per 10,000 manufacturing employees, compared to 152 globally on average (IFR, World Robotics 2026).
- According to Daron Acemoglu and Pascual Restrepo, automation can dissipate wage rents and amplify wage losses in targeted tasks.
- Germany, half as roboticized as South Korea, has maintained more stable wage progression, but the causal role of its codetermination mechanisms in this stability is not directly established.
- By 2040, the capacity of collective bargaining institutions to capture a share of gains produced by machines that workers do not own remains uncertain.
1,066 versus 152: What the gap really measures
The 2026 IFR ranking places South Korea first, ahead of Singapore and Germany. Singapore comes second with 873 robots per 10,000 employees, Germany third with 469. The global average reached a record 152 robots per 10,000 manufacturing employees. The gap between Seoul and the rest of the world is not measured in percentage but in multiples: seven times.
This density was built over two decades of industrial investment concentrated in a few sectors. Electronics and automobiles account for most of South Korea’s robot fleet. Samsung, Hyundai, LG mechanized their production lines at a pace few economies have matched. The chaebol, the large family conglomerate that structures Korean industry, had the means to buy the equipment and the expertise to integrate it.
The problem does not come from automation itself. South Korean manufacturing productivity advanced. Exports held up. Industry did not collapse. But the progression of real wages in South Korea’s manufacturing sector did not follow the productivity curve.
The OECD documents the acceleration of robotization, but not directly a Korean gap between robots and wages over this period.
In this study, robotization is associated with effects on employment that vary by profession and country, with declines particularly observed in certain intermediate and low-skilled professions. It has not necessarily produced better wages.
The dissipation thesis: when the machine eats the wage rent
Daron Acemoglu and Pascual Restrepo developed an analysis that directly illuminates what the South Korean figures show. Their central argument: automation can alter bargaining power and wage rents, depending on labor markets and companies. Workers who occupied positions where their scarcity or collective organization gave them bargaining power lose that leverage when the machine does the same work. The dissipation of rents linked to automation is one possible mechanism for divergence between productivity and wages, among other determinants.
This dynamic has a counterintuitive consequence. Even automation that increases overall productivity can be inefficient from a distribution perspective. If gains go entirely to equipment owners and displaced workers find lower-paid jobs, society produces more but distributes less equitably. GDP rises, so do inequalities.
South Korea offers a textbook case to test this thesis. South Korean collective bargaining is primarily decentralized and its coverage is limited, creating an institutional environment poorly equipped to capture automation gains for workers’ benefit.
This diagnosis is not without nuance. A competing reading emphasizes long-term effects. Automation destroys routine jobs but creates others, often better paid, in maintenance, supervision, and design. Over twenty years, the South Korean economy did indeed produce more skilled jobs. The question is whether this creation compensates for wage dissipation in automated tasks, and for whom.
Available data suggest compensation is partial and unequal. Skilled engineers and technicians benefited from industrial upgrading. Low-skilled operators, often employed in subcontracting, absorbed downward pressure. The employment structure changed in South Korean manufacturing, even if total employment did not necessarily decline.
The German codetermination model
Germany roboticizes less than South Korea, 469 versus 1,066. This difference does not stem from technology: the same machines, often the same equipment makers. Institutions constitute one possible explanatory factor among others, not a cause demonstrated by itself alone.
The German codetermination model gives worker representatives a seat on the supervisory board of large companies. Works councils negotiate the conditions for introducing new technologies. Collective agreements cover a large portion of the manufacturing sector and include clauses on the distribution of productivity gains. When BMW or Volkswagen installs a new robotic line, unions are at the bargaining table over what that means for wages, hours, and retraining.
Codetermination can influence wage distribution and performance, but its role in capturing automation gains is not demonstrated here. It did not block automation; Germany remains one of the world’s most roboticized economies.
This model has its limits. It applies to large companies and workers on permanent contracts. Subcontractors, temporary workers, and SMEs are often outside the scope of codetermination. And competitive pressure from low-cost economies has eroded negotiating margins in some sectors. But compared to South Korea, the institutional gap remains striking.
Institutions can be an explanatory hypothesis for the German contrast, but the sources consulted do not allow making it an established conclusion of a German-South Korean comparison. Robot density does not mechanically determine wage trajectory. Negotiation and sharing mechanisms influence the distribution of gains, among other wage determinants.
Singapore, the third case: performance without visible redistribution
Singapore complicates the picture. With 873 robots per 10,000 employees, the city-state ranks just behind South Korea. Its economy is prosperous, its wages are high in absolute terms, and its unemployment rate remains low. The Singapore model seems to contradict the dissipation thesis: high robotization, good wages, employment preserved.
Two elements nuance this picture. First, Singapore is a service economy as much as an industrial one: its robot density reflects a highly concentrated and high-end manufacturing sector, not mass industry. Comparisons with South Korea, which has a much broader manufacturing base, are therefore delicate. Second, Singapore maintains notable salary and income inequalities, but official data do not place it among the most unequal developed economies. Prosperity is real, but its distribution is highly concentrated.
Singapore is therefore not a counterexample to Johnson’s thesis. It is rather a case where high productivity compensates in terms of living standards what institutional weakness leaves intact in terms of distribution. Singapore workers earn well, but they would earn better if automation gains were shared more equitably. This is a discreet form of dissipation, less visible than in South Korea because it occurs at a higher absolute level.
Machine ownership and value sharing by 2040
The question looming over the next fifteen years is more complex than robot density. Industrial robots, until now reserved for large groups, are becoming accessible to medium-sized companies. External projection anticipated a drop of around 65% in the unit cost of industrial robots, down to less than $11,000 by 2025, corresponding to roughly a threefold decrease since 2014. This diffusion beyond chaebols and major automakers raises a new institutional question.
When a large South Korean group roboticizes, it is visible, negotiable, subject to union pressure and public attention. When thousands of SMEs each roboticize at their own scale, the dissipation of wage rents occurs in diffuse fashion, off the radar of collective bargaining. Sector-wide agreements may lose their bite if the companies signing them represent a declining share of employment.
Two trajectories are plausible by 2040. In the first, the codetermination mechanisms documented in Germany extend to economies like South Korea, either through institutional borrowing or under political pressure from a working class impoverished by dissipation. Signals in this direction exist: South Korean unions have achieved advances on automation investment transparency in several sector agreements since 2022.
In the second trajectory, the diffusion of automation to SMEs bypasses existing collective rights. Workers in the most automated companies lose their bargaining power before institutions have time to adapt. Gains remain captured by equipment owners, and income inequalities in the manufacturing sector widen durably. The article published in this journal on the concentration of new jobs in low-wage sectors documents a similar dynamic for Western economies.
What would allow distinguishing between the two trajectories is observable right now: the evolution of wages’ share in manufacturing value added in the most roboticized economies, and the coverage rate by collective agreements in sectors with high robot density. If South Korea’s wage share continues to fall while robot density rises, dissipation accelerates. If sector agreements including automation clauses multiply, institutional capture progresses.
One concrete avenue is being discussed in several industrial policy circles: conditioning tax advantages tied to automation investment on commitments to training and wage bargaining. The logic is straightforward. The state that partially finances robot purchase through tax deductions can in return require that productivity gains do not evaporate entirely into margins. This does not resolve the question of machine ownership, but it creates an indirect redistribution mechanism. Devices in this spirit already exist in Germany and Austria, linked to industrial modernization aid.
The article on supporting SMEs in the face of AI gains emphasizes that technological diffusion without institutional support widens gaps between companies. The same reasoning applies to physical robotics: without a negotiated framework, diffusion to SMEs risks producing diffuse wage dissipation that is difficult to correct after the fact.
The limits of robot density as an indicator
The IFR ranking provides a precise image of a partial reality. South Korea is indeed seven times more roboticized than the global average. This density reflects real industrial performance, massive investment in productive modernization, capacity to compete on demanding export markets.
It does not inform on South Korean workers’ association with gains produced by their factories, nor on Germany’s capacity, less roboticized, to better distribute machine benefits. These elements appear in institutions, collective agreements, ownership structures, and retraining policies.
The challenge of the next fifteen years is less about whether automation continues—it will—than about whether societies accelerating it equip themselves with institutions to distribute its fruits. Germany built these institutions before robot density became a problem. South Korea is building its own as catch-up, with an institutional gap to close. And for economies automating today with weak union coverage, the window to act closes as machine ownership concentrates.
Sources
- International Federation of Robotics, World Robotics 2026
- Simon Johnson & Daron Acemoglu, Automation and Rent Dissipation: Implications for Wages, Inequality, and Productivity, MIT Economics
- OECD, work on automation, productivity and income distribution (available at stats.oecd.org)



