Between 2018 and 2022, in five ASEAN countries, estimates attribute to robotization more creation of formal skilled jobs than displacement of formal low-skilled jobs, while results for developed economies varied by country and period, reflecting robotic deployments whose exposure, sectoral structures, and adoption levels differed. The outcome depends on both the characteristics of the technology and the economic context of its adoption. For workers in middle-income economies potentially affected by automation by 2035, this distinction changes everything.
The essentials
- Robotization can create or displace jobs depending on productivity, scale, reallocation effects and economic conditions, not on machine density alone.
- Between 2018 and 2022, in five ASEAN countries, estimates attribute to robotization more creation of formal skilled jobs than displacement of formal low-skilled jobs (World Bank, Future Jobs 2025).
- In North America and Europe, employment balances between 2018 and 2022 varied by country and period.
- Daron Acemoglu, 2024 Nobel Prize winner in economics, argues that it is institutions, retraining policies, collective bargaining, and economic regimes that distribute productivity gains, not technology alone.
- Europe has a window to reorient its trajectory, provided it links robotic investment to active employment policies.
Vietnam and Germany deployed the same robots, with opposite results
Vietnam increased its imports of industrial robots between 2018 and 2022. Between 2018 and 2022, the value of German imports of industrial robots declined, despite annual variations, on an already much denser fleet. In both cases, production lines were automated, data entry and assembly operators were replaced. Yet Vietnam recorded net job creation; some German studies find negative local effects from robotization, but the report does not conclude with a net job destruction in Germany.
The explanation lies in simple mechanics. Growth and economies of scale can promote the absorption of displaced workers, without guaranteeing it for everyone. The business services sector, logistics, maintenance of automated equipment, commercial and technical functions: several of these segments developed rapidly during this period. The economy acts like a sponge. In Germany, this sponge is saturated: the productive base is mature, growth is weak, and each position eliminated by a machine does not automatically find its equivalent elsewhere.
World Bank data highlight the mechanism in Vietnam and extrapolate the orders of magnitude to ASEAN-5. Over the period, a significant number of low-skilled workers were displaced by robots in the formal sector. But more higher-skilled positions were created in the same economies. The balance is positive. In North America and Europe, employment balances differed depending on national contexts and periods.
European robotization was smaller in volume than that of Asia, but more dynamic than that of the Americas over 2018-2022. The absorption of displaced workers varied by country and period in Europe compared to Southeast Asia.
The role of institutions where technology is not enough
Daron Acemoglu, who received the 2024 Nobel Prize in economics for his work on institutions and prosperity, emphasizes that technology does not distribute its own gains. It is institutions and the choice of technological trajectory that influence the distribution of productivity gains.
Applied to robotization, this framework directly illuminates the observed asymmetry. Southeast Asian countries deployed robots within a framework of sustained growth, with active policies: retraining programs targeted at expanding sectors, facilitated job mobility, public investment in infrastructure that creates new jobs. Productivity gains were associated with increased production scale, which could offset some of the displacements.
In Europe, the situation is more mixed. Social protection systems cushion the shock for individuals, but available assessments cited show that certain targeted training or retraining programs can mitigate adverse effects and facilitate transitions. The OECD’s 2026 Employment Outlook report addresses geographic disparities, skills, and training in particular. The least qualified workers, first displaced by robots, have unequal access to existing programs.
Acemoglu’s argument goes further: a technology deployed without adapted regulation tends to concentrate gains among those who control it. Europe is falling behind on robots not only in equipment density, but also in capacity to build the institutional framework that distributes the gains from automation.
The competing reading: growth alone is not enough
Another reading of the same phenomenon, less institutionalist, deserves consideration. Tyler Cowen and economists close to progress studies argue that economic growth is itself the product of supply-side policies, and not an exogenous given on which institutions would graft themselves. In this perspective, Europe’s problem is not primarily institutional: it is a growth deficit rooted in structural rigidities, excessive regulation, fragmented markets, high labor taxation, which reduces the absorption capacity of economies before robotization even comes into play.
This reading points toward different prescriptions. Reform first to restart growth, and jobs will follow, including in automated sectors. Retraining remains necessary, but it is only one tool among others, secondary to the macroeconomic environment.
World Bank data do not constitute a formal test between these two hypotheses. Southeast Asia indeed combines strong growth and active policies. The report mentions more favorable European results in countries with low or moderate labor costs, without isolating the effect of training.
The truth is probably that both factors interact: growth widens the sponge, institutions determine who accesses new positions and within what timeframe. Adopting AI without training locals amounts to outsourcing growth: this logic also applies to industrial robotization.
The 85 million workers of 2035 by their country
In 2023, China’s fifth economic census counts 104.8 million people employed in manufacturing by industrial enterprises, while 85 million corresponds to an earlier World Bank estimate. From 2022, the World Bank projects that at the rate observed between 2018 and 2022, Vietnam could reach in approximately 13 years the current robotization level of high-income economies. This figure is often presented as a threat. It can also be read as a conditional opportunity.
The data suggest a role for productivity gains and scale; they do not permit attribution of the net balance to retraining policies. In the ASEAN-5 studied, estimated creations concern formal workers with tertiary education, without corresponding sectoral breakdown.
But several conditions must be met. The first is that growth remains strong enough to generate replacement jobs within humanly acceptable timeframes. A decade of transition is a decade of unemployment for those affected, even if the final balance is positive. The second is that educational systems produce skills adapted to the positions created. In several economies, vocational training capacity remains limited in the face of expected transitions.
Acemoglu emphasizes a third condition: the need to ensure equitable distribution of productivity gains generated by automation. In economies where collective bargaining is weak and labor regulation insufficient, nothing mechanically ensures this sharing.
The requirements of the 2035 horizon for mature economies
For Europe and North America, the 2035 horizon poses a challenge: improve the employment balance in the face of automation without depending on a return to strong growth.
Several levers exist, documented by the OECD. The first is to condition public aid for robotic investment on retraining commitments. Several European countries have experimented with this logic with measurable results on the retention of displaced workers in the formal labor market. The second is to invest in sectors with high potential for qualified job creation—energy transition, care for persons, maintenance of digital infrastructure—which absorb intermediate skills accessible to workers displaced by industrial automation.
The third lever, the most debated, is direct regulation of the pace of automation. Acemoglu has proposed studying a robot tax that would rebalance the price signal between labor and automated capital, currently judged too favorable to rapid automation. The idea remains controversial: its liberal opponents argue it would slow adoption of genuine productivity gains and discourage investment. The debate is open, and available data do not yet permit a conclusion.
Without appropriate policies, the risk of displacement and inequality increases in low-growth economies. The question is whether European institutions will be redesigned to distribute the gains from automation more broadly than markets alone do.
Two institutional regimes facing the same machine
The 2018-2022 asymmetry does not say that Southeast Asia managed robotization better than Europe. It says that two different economic regimes produce different effects with the same tools. An expanding economy creates more spaces for employment to absorb technological transitions.
For European policymakers, restarting growth and reforming institutions for sharing gains are two complementary agendas. The first without the second produces growth that concentrates benefits. The second without the first results in redistribution without sufficient productive base to finance it.
The World Bank analyzes potential exposures to AI with earlier data, but does not document this claim for 2026: in several regions, positions created could eventually be exposed to automation as AI models advance. An acceleration of this movement could call into question some of the job gains observed between 2018 and 2022. This signal, for now weak, deserves careful monitoring by anyone wanting to extrapolate trends from the last decade to the 2035 horizon.
The World Bank and OECD have documented it: the social balance of automation depends on the economy in which the robot is inserted and the rules that societies set for themselves to distribute what it produces.
Sources
- World Bank, Future of Jobs Report 2026, documents1.worldbank.org
- OECD, Employment Outlook 2026, OECD Publishing (link not guaranteed)
- Daron Acemoglu, Nobel Lecture: Institutions, Technology and Prosperity (2024), nobelprize.org
- Acemoglu, D. & Restrepo, P., Robots and Employment: Evidence from Europe (2020-2022), American Economic Review (link not guaranteed)
