GLOBALISE GOODS, SELECT PEOPLE, PROTECT CAPABILITIES

The case for a Productive Citizen Economy, and the places where it breaks

Article 2 of 4  ·  Series I  ·  Published 9 October 2026  ·  Analysis → Forecast → Recommendations

How this series measures things. Every article applies the same three questions to its subject. Core size: how many residents at the same time work, earn above subsistence, pay more in tax than they receive, and spend broadly at home. Core productivity: how much value each of them produces, and how much capital, automation and AI stands behind each one. Core capture: how much of that value stays inside the country as wages, taxes and domestic demand. Article 1 showed that the Core is shrinking faster than the population. This article asks which production model can raise all three measures at once, and what it costs.

1. The Signal

On 4 March 2026 the European Commission published a proposal that would have been hard to imagine in Brussels ten years ago. The Industrial Accelerator Act sets a goal of lifting manufacturing to 20% of EU GDP by 2035. Today it is 15.9% of gross value added (measured, Eurostat 2024).

The proposal does more than set a target. It writes origin rules into public money. Solar installations that receive public funding would need an EU-made inverter and at least two other main components made in the EU within a year of the law taking effect, and the inverter, the cell and at least four EU-made components within three years. Concrete, mortar and aluminium bought with public funds would face a mix of low-carbon and origin requirements. Foreign investments above €100 million in batteries, electric vehicles, solar and critical raw materials would have to meet at least four of six conditions. Among them: ownership of an EU company capped at 49%, at least 1% of revenue spent on EU research every year, at least half the workforce made up of EU workers, and an effort to source at least 30% of inputs inside the EU (reported, COM(2026) 100, via Mayer Brown).

Investors from countries with free-trade or customs-union agreements are exempt. Authorities may waive the origin rules if they delay a project by more than seven months or raise its cost by more than 30%. This is not autarky. It is something more specific: open for most trade, conditional for a short list of strategic industries, and explicit that jobs and know-how are part of the price.

The timing is not accidental. In 2025 German industry cut about 124,000 jobs, a fall of 2.3% to roughly 5.38 million. The car industry alone lost about 50,000 positions in a year and 111,000 since 2019, a drop of 13% (reported, EY analysis of Destatis data). Article 1 explained part of the reason: German exports to China fell 23% from their 2022 peak, and car exports to China fell by about two thirds. A former customer has become a competitor.

So Europe's largest economy is losing industrial jobs at a pace of about 10,000 a month (derived), at the same time as its working-age population begins a decade of decline. The Commission's answer is a deliberate break with three decades of trade policy built on the idea that the location of production does not matter.

The question this article addresses is whether that answer is right, and if so, in what form. Our view is that the direction is right and the instruments are fragile. The direction follows from demographics. The fragility follows from history.

2. The Mechanism: why a shrinking Core changes the price of a lost job

For thirty years the case for open markets rested on one transaction, and on its own terms the transaction is correct. A television made abroad for €300 instead of €600 at home saves the buyer €300. Multiply that across millions of purchases and an economy becomes richer in consumer goods, and the workers who lose their jobs in the television industry move to other work.

The difficulty lies in the last clause. The theory assumed that displaced workers find equivalent jobs quickly. In much of the rich world, they did not.

What the China shock taught

The most influential research on this question comes from David Autor, David Dorn and Gordon Hanson, and colleagues. Studying the period after China joined the World Trade Organization, they found that rising Chinese import competition between 1999 and 2011 accounted for roughly 2.0 to 2.4 million lost US jobs (measured, Acemoglu, Autor, Dorn, Hanson and Price, Journal of Labor Economics 2016). The damage did not spread evenly and fade away. It concentrated in specific places, where employment, participation and earnings stayed depressed for more than a decade. Workers did not move in large numbers to booming regions. Many left the workforce.

That finding did not overturn the case for trade. It overturned the assumption that the costs were small and temporary.

Now add ageing

The China shock hit the United States when its working-age population was still growing. Young cohorts kept arriving, and over time new sectors absorbed them even when the displaced older workers did not recover.

An ageing economy lacks that reserve. When Germany loses a 40-year-old engineer earning €80,000, it loses a wage, the income tax and social contributions on it, the household spending it finances, the mortgage the household carries, and the savings it adds to the pension system. Germany's tax wedge on a single worker at the average wage is 49.3% of total labour cost, the second highest in the OECD after Belgium and well above the OECD average of 35.1% (measured, OECD Taxing Wages 2026). Roughly half of what an employer pays for that engineer comes back to the state.

The pool from which a replacement could come is shrinking. Germany's population aged 20 to 66 is projected to fall by between 3.2 and 6.2 million by 2035, depending on migration (projected, Destatis). The Draghi report on European competitiveness put it bluntly: by 2040 the EU workforce is projected to shrink by close to two million workers a year (reported, Draghi 2024).

The implication is narrow but important. The social cost of destroying a good job rises as demographics worsen. This is not an argument against trade. It changes which trade-offs are worth making.

The total social cost of an import

When a firm decides where to produce, it compares private costs: labour, energy, logistics, tariffs. When a country decides which industries to support, the relevant comparison is different. Call it the total social cost of an import:

—   the price difference between the imported and the domestic product,

—   minus the domestic wages and taxes that local production would generate,

—   minus the spillovers of local production: engineers trained, suppliers sustained, process knowledge kept, firms started by former employees,

—   minus the value of resilience: what it costs when a single-source supply is cut,

—   plus the efficiency loss of producing at home what could be made better elsewhere.

For most products the first and last terms dominate and trade wins easily. For a small set of products, the middle terms are large and grow larger as the Core shrinks.

Three classes of goods

That leads to a practical sorting, which is what the Industrial Accelerator Act is attempting in legal form.

The middle row is short as a list of products. It is not short as a share of what matters for the Core: it holds a disproportionate part of high-wage industrial employment, engineering talent and business research spending.

Why cheap imports matter less as a country gets richer

One more mechanism supports the thesis. For a household earning €15,000 a year, a €300 saving on a television is significant. For an engineer earning €100,000, it is marginal, while losing the €100,000 job would be catastrophic for the household and a serious loss for the state. A modest price premium on strategic goods can be a rational insurance payment for keeping a high-productivity economy intact.

The word that matters is modest. If domestic goods cost three times as much, the policy destroys the purchasing power it was meant to protect. The Commission's own 30% cost threshold for waiving origin rules is, in effect, an estimate of where that line sits.

The thesis in one paragraph

When working-age populations shrink, the scarcest national asset is no longer capital or factory floor space. It is the resident who, at the same time, produces high value, earns a high wage, pays substantial tax and spends broadly at home. Economic strategy should therefore aim to maximise the number of such people and the output of each one. That means pairing AI and robotics with expensive, skilled workers rather than with cheap labour; admitting migrants selectively and converting them quickly into self-sufficient taxpayers; and choosing a deliberate middle point between closure and hyper-globalisation, where ordinary goods trade freely but strategic capabilities stay at home or among close partners. We call this the Productive Citizen Economy.

3. Subtheme One: automation that multiplies the expensive worker

There are two ways to automate a factory, and they lead to different societies.

In the first, the machine replaces the worker. A hundred jobs at €50,000 become ten engineering jobs at €150,000 and ninety people pushed into lower-paid services. Profit can rise while the base of taxpayers and consumers erodes.

In the second, the machine makes the expensive worker economically viable. A hundred jobs at €50,000 become sixty jobs at €80,000, each supported by robots, sensors and AI, while the other forty positions disappear gradually through retirement rather than dismissal. Output per worker climbs, pay climbs, tax per worker climbs.

Both happen in practice. Which one dominates is the central question of this subtheme.

What the evidence shows about robots and jobs

Daron Acemoglu and Pascual Restrepo studied US labour markets exposed to industrial robots. Each additional robot per thousand workers reduced the employment-to-population ratio by about 0.2 percentage points and wages by about 0.4% (measured, Journal of Political Economy 2020). In the United States, robots were largely the first kind of automation. Acemoglu and Restrepo coined a term for technologies that displace workers without delivering large productivity gains: so-so automation. Self-checkout machines are the familiar example. They shift work to the customer and cut payroll without making the shop much more productive.

Germany shows a different pattern. Wolfgang Dauth, Sebastian Findeisen, Jens Suedekum and Nicole Woessner found that robot adoption in Germany between 1994 and 2014 did not reduce total employment. Each robot was associated with about two fewer manufacturing jobs, but these were mostly jobs that were never created for young entrants rather than existing workers dismissed, and the losses were offset by new service jobs. Incumbent workers in exposed industries tended to stay with their employer and move into different tasks, often with higher qualification requirements, although wage growth for some middle-skilled groups was weaker (measured, Journal of the European Economic Association 2021).

The contrast is instructive. The same machines produced different outcomes because the surrounding institutions differed: apprenticeship systems that retrain, works councils that negotiate redeployment, firms that held on to skilled staff because hiring them back would be expensive.

Labour scarcity pushes automation toward the better kind

Ageing economies have an advantage here that is rarely noticed. When there is nobody to hire, a robot does not replace a worker who wanted the job. It fills a vacancy that would otherwise stay empty, and the people who remain become more valuable.

The data on robot density, the number of industrial robots per 10,000 manufacturing employees, follows demographics more closely than wage levels.

Source: International Federation of Robotics, World Robotics 2024 (measured).

The fastest-ageing economies automate the most. Acemoglu and Restrepo themselves found this across countries, and argued it explains why ageing has not, on average, reduced growth in GDP per capita: ageing economies compensate by investing in machines (Acemoglu and Restrepo, AER Papers and Proceedings 2017).

Japan shows the mechanism outside the factory as well. Workers aged 65 and over are 13.8% of all employed people (measured, 2025). Restaurants, retailers, logistics firms and builders are investing in labour-saving equipment because hiring is not available at any reasonable price.

The loop worth aiming for looks like this:

fewer workers → labour scarcity → higher wages → automation pays → more capital per worker → higher output per worker → higher wages, tax and spending per worker

The loop worth avoiding is its mirror image:

fewer workers → large-scale import of low-wage labour → wages held down → weaker incentive to automate → productivity stagnates → the Core thins in both size and dynamism

AI changes who gets multiplied

Industrial robots mostly multiply people who work with physical processes. AI extends the same logic to people who work with information: the engineer who designs with simulation tools, the maintenance technician who diagnoses with sensor data, the nurse whose documentation is drafted automatically, the accountant who supervises software rather than entering data.

The early evidence on generative AI at work points the same way. In a large study of customer-support agents, an AI assistant raised productivity by about 14% on average and by about a third for the least experienced, because it spread the know-how of the best performers (measured, Brynjolfsson, Li and Raymond, Quarterly Journal of Economics 2025). If that holds, AI could help an ageing economy bring new and migrant workers up to full productivity faster, which matters directly for the second subtheme.

But AI does not choose its own deployment. A firm can use the same model to raise each employee's output or to cut the number of employees. In a labour-scarce economy the first choice is more often the profitable one. In an economy with abundant cheap labour, the second can win. Demography tilts the choice; it does not make it.

Where the automation thesis breaks

The strongest counter-example comes from a company that tried exactly this strategy. In 2016 Adidas opened a highly automated "Speedfactory" in Ansbach, Germany, followed by a second in Atlanta, to produce trainers close to their customers. By 2020 both had closed and the technology had moved to suppliers in Asia. Automation worked in the narrow sense. What failed was the ecosystem: the component suppliers, the flexibility of large contract manufacturers, and the cost of producing a small range of models at scale in high-cost locations.

The lesson is not that reshoring through automation cannot work. It is that a single automated plant cannot substitute for a supplier network. Strategic ecosystems, the middle row of the table in Section 2, are worth defending precisely because they are hard to rebuild one factory at a time.

4. Subtheme Two: migration as conversion, not headcount

In population statistics every arriving adult counts as one. In a Productive Citizen Economy, what matters is the path from arrival to economic self-sufficiency, and how long it takes.

Three kinds of contribution

Three groups add to the Core, not one.

—   High human capital: engineers, doctors, researchers, founders. Research by Shai Bernstein and co-authors found that immigrants made up about 16% of US inventors but produced about 23% of patented innovation, with additional gains through their effect on native colleagues (measured, NBER 2022).

—   Scarce skilled hands: electricians, welders, technicians, nurses, mechanics, builders. Ageing societies need these at least as much as software engineers, and the infrastructure backlog described in Article 1 depends on them.

—   High employability: people without a rare qualification but with the language, reliability and drive to enter work quickly and move up.

The evidence on lower-skilled migration is genuinely divided. Some studies find small negative wage effects for the lowest-paid native workers; others find neutral or slightly positive effects as natives and newcomers specialise in different tasks. Ageing societies also need care workers, drivers, builders and agricultural labour in areas where automation advances slowly. A realistic policy is selective and adaptive, not limited to graduates.

Conversion speed decides the outcome

Germany illustrates why speed matters more than numbers. The OECD measured a migrant–native employment gap of 10.3 percentage points in Germany in 2024, among the widest in the organisation (measured, OECD International Migration Outlook 2025). Yet the Institute for Employment Research found that 64% of refugees who arrived in 2015 were employed by 2024, approaching the national rate of about 70% (reported, IAB). Both facts are true. Conversion happens, but slowly.

A new resident who reaches self-sufficiency in two years is a demographic asset. One who takes ten years is a fiscal cost for most of that decade. The difference lies less in who arrives than in what happens after arrival: recognition of qualifications, language training tied to work, employer involvement, housing near jobs, and childcare that allows both partners to work.

What selection looks like in practice

Germany introduced an Opportunity Card in June 2024, a points-based visa allowing skilled workers to come and look for a job for up to a year. By mid-June 2025, German missions abroad had issued 11,497 of them, almost a third to Indian citizens (reported, German Economic Institute). That is a useful start and a small number. Destatis estimates that holding the 20-to-66 population steady to 2035 would require about 570,000 working-age immigrants every year. The gap between the two figures shows how far selective migration has to scale before it shapes the Core rather than individual firms.

Japan has moved faster from a lower base. Foreign workers reached a record 2.57 million in October 2025, up 11.7% in a year. Vietnam is the largest source, followed by China and the Philippines (measured, Ministry of Health, Labour and Welfare via Nippon.com). A country long considered closed to immigration is now adding several hundred thousand foreign workers a year, largely in manufacturing, care, construction and services, through programmes that tie residence to specific skills and sectors.

Integration defined by what can be measured

Integration is often discussed in terms of culture and values. Those matter, but they are hard to measure and easy to politicise. A Productive Citizen Economy needs a definition it can track and act on:

—   language sufficient for the job and for daily life,

—   employment, and time taken to reach it,

—   economic self-sufficiency: earning more than the household receives in transfers,

—   children's educational outcomes,

—   respect for the law,

—   long-term participation in the country's institutions.

Each of these can be counted. Countries that measure them can see which programmes convert newcomers quickly and which do not, and fund the first. Shared principles tend to follow shared work, schools and neighbourhoods. They rarely precede them.

5. What Most Analysis Gets Wrong

The false binary

Public debate usually presents a choice between free trade and protectionism. Neither describes what ageing economies are actually doing. The United States, China, Japan, Korea and now the EU are all converging on the same hybrid: open markets for most goods, conditional access for a short list of strategic sectors, and active industrial policy within that list. The relevant question is no longer whether to have industrial policy but how to stop it becoming permanent subsidy.

Counting jobs instead of the Core

Most assessments of trade and industrial policy count jobs created or saved. That is the wrong metric for an ageing economy. A subsidised plant that employs 3,000 people at average wages, poaching them from other firms in a tight labour market, may add little to the Core. A plant that employs 600 people at high wages, trains them, and supports a network of suppliers may add much more. What matters is the effect on Core size, Core productivity and Core capture, not headcount on the factory floor.

Ignoring the failure record

The case for strategic industrial policy is strong. The record of individual projects is mixed. In July 2025 Intel cancelled its planned €30 billion chip factory in Magdeburg, which had been promised €10 billion of German public money and 3,000 jobs; the subsidy had not yet been paid (reported, Brussels Signal). In March 2025 Northvolt, Europe's flagship battery manufacturer, filed for bankruptcy in Sweden despite substantial public and private backing. Analyses that cite only the strategic logic, and not this record, overstate what governments can engineer.

The IMF has made the point carefully. Well-targeted industrial policy can raise productivity where learning-by-doing and scale effects are strong and where governments have the capacity to select and monitor projects. Mis-targeted policy raises prices, misallocates capital and labour, and invites retaliation. The difference is not ideological. It is a matter of design and discipline.

Overlooking who pays

Higher prices on protected goods fall hardest on lower-income households, who spend a larger share of their income on goods. If protection spreads from strategic ecosystems into ordinary consumer products, the policy becomes regressive and erodes the purchasing power of the Core it was meant to sustain.

Where the thesis sits among the economists

Each school holds a piece. The combination, demographic contraction, scarcity of productive consumers, the rising value of each good job, AI and robots as multipliers of expensive workers, selective migration, and strategic openness, is less often stated as one system.

6. Base / Stress / Extreme

The scenarios describe the EU, Japan and Korea over the next five years. They sum to 100.

Base: Strategic openness, unevenly executed (50%). The Industrial Accelerator Act is adopted in a diluted form, with origin rules narrowed and waivers used often. Strategic sectors receive targeted support with mixed results: some reshoring succeeds where supplier networks exist, several large subsidised projects are cancelled or delayed. Robot density keeps rising. Skilled migration expands, but conversion into self-sufficient taxpayers remains slow in Germany and faster in Japan. Industrial employment continues to fall gently while output per worker rises.

Falsifier: EU manufacturing value added falls by more than 5% in real terms between 2025 and 2029 while EU robot density stagnates.

Stress: Protection without productivity (35%). Origin rules and subsidies spread beyond the strategic list into politically sensitive sectors. Prices rise, competition weakens, and subsidised plants underperform. High energy costs keep European industry uncompetitive regardless of support. Labour scarcity is met mainly with low-wage temporary migration rather than automation, and the labour share falls. Consumers pay more for goods while the Core still shrinks.

Falsifier: output per hour in German manufacturing grows above 2% a year over 2026–2029.

Extreme: Closure spiral (15%). Trade conflict between the EU and China escalates into broad tariffs and export controls on critical inputs, including rare earths and battery materials, of the kind China applied in April 2025. Input costs jump, retaliation hits European exporters, and several strategic supply chains face physical shortages for months. Protection becomes reactive rather than selective.

Falsifier: the share of EU–China goods trade covered by new tariffs or export restrictions stays below 20% through 2029.

7. Forecast, 1 year (to October 2027)

F1. The Council and the European Parliament announce a provisional political agreement on the Industrial Accelerator Act by 30 September 2027.

Comparable files moved quickly: the Net-Zero Industry Act went from proposal in March 2023 to provisional agreement in February 2024, and the Critical Raw Materials Act from March 2023 to November 2023. The IAA is more contested because origin rules divide member states with large and small manufacturing bases, and because FDI conditions affect investors from Asia and the Gulf.

Probability: 0.55. Confidence: Medium.

Second-order effect: firms in solar, batteries, automotive and construction materials begin redesigning supply chains for public contracts before the law applies.

Weakening condition: a major trade agreement or a ceasefire in trade tensions with China that reduces political urgency, or a member-state coalition blocking origin rules in Council.

Wider 1-year view: expect further industrial job losses in Germany at a pace close to 2025, more announcements of automation investment explicitly justified by labour shortages, and continued double-digit growth in Japan's foreign workforce.

8. Forecast, 3 years (to 2029)

F2. The International Federation of Robotics reports Germany's robot density for 2028 at or above 500 robots per 10,000 manufacturing employees.

Germany's density was 429 in 2023, growing about 5% a year since 2018 (measured, IFR). Five years at that pace would reach about 547 (derived). Two factors push in the same direction: the denominator, manufacturing employment, is falling, and labour scarcity raises the return on automation.

Probability: 0.65. Confidence: Medium.

Second-order effect: output per worker in German manufacturing rises even as headcount falls, making industrial wages one of the few areas where real pay keeps pace with productivity.

Weakening condition: a deep and prolonged investment slump in German industry, or a large-scale shift of German production capacity abroad, which would cut robot installations along with jobs.

Wider 3-year view: at least one large European manufacturer announces a new highly automated domestic plant explicitly justified by proximity to EU customers and public-procurement origin rules; at least one large subsidised strategic project is cancelled or substantially scaled back.

9. Forecast, 5 years (to 2031)

F3. Japan's Ministry of Health, Labour and Welfare reports 4.0 million or more foreign workers as of the end of October 2030.

The count was 2.57 million in October 2025, up 11.7% in a year (measured). Growth at 9.5% a year would reach 4.0 million by 2030 (derived). The government has expanded the sectors open to specified skilled workers and is replacing the technical intern programme with a scheme designed to retain and upgrade workers. Labour shortages in care, construction, logistics and manufacturing continue to deepen.

Probability: 0.55. Confidence: Medium-Low.

Second-order effect: Japan becomes the clearest test of whether selective, sector-tied migration can raise the Core in a society with little immigration history.

Weakening condition: political pressure to cap foreign residents, a sharp yen depreciation that makes Japan less attractive relative to Korea, Taiwan and Australia, or a recession that reduces vacancies.

Wider 5-year view: a widening gap between Japan's and Germany's migrant conversion rates becomes visible in employment data, and policy debate in Europe shifts from numbers of arrivals to time to self-sufficiency.

10. Forecast, 10 years (to 2036)

F4. Eurostat reports manufacturing at 18.0% or more of EU-27 gross value added for the year 2034.

The share was 15.9% in 2024 (measured). The Commission's target is 20% of GDP by 2035. Reaching even 18% would require manufacturing to grow roughly 1.2 percentage points a year faster than the rest of the economy for a decade (derived). Europe faces higher energy costs than its main competitors, an investment gap estimated by the Draghi report at €750–800 billion a year, and a shrinking workforce.

Probability: 0.15. Confidence: Medium.

Second-order effect: if the target is missed, as we expect, strategic industrial policy will be judged by the wrong yardstick. Higher output per worker, stronger strategic capabilities and a stable high-wage industrial Core are compatible with a flat or falling manufacturing share.

Weakening condition: a large fall in European energy prices, a major defence-industrial expansion that counts toward manufacturing value added, or a sharp contraction in services value added.

This forecast is written deliberately against the official target. Its purpose in the record is to test whether the system reads the limits of industrial policy correctly.

11. Signals to Watch

1. Progress of the Industrial Accelerator Act in Council and Parliament, and especially whether origin rules are narrowed or widened.

2. Use of waivers under existing origin rules: how often authorities cite delays above seven months or cost increases above 30%.

3. Robot density and installations in Germany, Japan, Korea and China, from the IFR's annual report.

4. Output per hour in manufacturing against industrial employment. Rising productivity with falling headcount is the Productive Citizen pattern; falling headcount without productivity gains is decline.

5. Labour share of income in the high-robot-density economies. If it falls, automation is not reaching wages.

6. Migrant conversion indicators: Germany's migrant–native employment gap and Japan's foreign worker count by visa type.

7. Cancelled or delayed subsidised projects in semiconductors, batteries and green steel, as a measure of execution risk.

12. Recommendations, Individuals

Immediate (within 90 days). List the three tasks that take most of your working week and test each against the AI tools now available in your field. Keep the tasks where the tools make you faster or more accurate; note the ones they could do without you. The first group is your future value; the second is your exposure.

Build (6–24 months). Move toward roles that combine domain judgement with the ability to direct automated systems: maintenance and inspection with sensor data, engineering with simulation, care coordination with digital records, finance and law with AI-assisted analysis. If you work in a strategic sector named in the Industrial Accelerator Act, such as batteries, solar, wind, grid equipment, automotive or construction materials, learn the origin and low-carbon requirements that will shape your employer's contracts; people who understand them will be needed. If you are considering working abroad, compare how quickly each destination recognises qualifications and how long its newcomers take to reach full employment, not only salary.

Position (2–10 years). Expect a modest premium on some strategic goods, including certain energy equipment, vehicles and their components, and plan large household purchases with that in mind. Favour careers and locations where your skills sit inside a supplier network rather than a single plant: networks survive downturns better than individual sites.

Avoid. Building a career on tasks that are routine, remote and easily priced against global labour or AI. Assuming that protection will keep any particular plant open: subsidies are withdrawn, projects are cancelled, and the Intel and Northvolt cases happened within months of each other.

Why. In a Productive Citizen Economy, value goes to people whom machines multiply. Policy can protect industries for a time; it cannot protect a skill set that has stopped being scarce.

13. Recommendations, Business

Immediate (this quarter). Classify your major inputs into the three classes: location irrelevant, strategic, domestic service. For every strategic input, record the share sourced from a single country and price a second source. If you sell into EU public procurement in solar, batteries, automotive, steel, aluminium, cement or concrete, map which of your products would meet the proposed origin and low-carbon requirements and which would not.

Build (6–24 months). Design automation and AI projects around output per employee and wages, not headcount reduction. In labour-scarce markets, the people who cannot be replaced are the ones who make automation work, and losing them costs more than the machines save. Pair every major automation investment with a retraining plan for the staff it affects. If you hire internationally, budget for qualification recognition, language training linked to the job, and help with housing; the speed at which a new hire reaches full productivity is the real cost driver. If you are a non-EU investor planning a manufacturing project above €100 million in batteries, electric vehicles, solar or critical raw materials, prepare now for the proposed conditions on ownership, research spending, EU workforce share and local sourcing.

Position (2–10 years). Re-evaluate production location using total cost, including logistics and trade-policy risk, loss of engineering knowledge and distance from customers. Where automation reduces labour to below roughly 15–20% of production cost, the case for low-wage locations weakens. Locate where a supplier network exists or can be built; the Speedfactory case shows that a single automated plant without one rarely survives. Treat public support as a bridge, not a business model, and plan for its withdrawal.

Avoid. Business cases that depend on subsidies or origin rules continuing indefinitely. Automation projects justified only by headcount cuts. Single-country dependence for strategic inputs, especially rare earths, power electronics and battery materials, where export controls have already been used.

Why. The policy environment is shifting from cost minimisation to capability preservation. Firms that combine automation with skilled, well-paid staff and resilient supply chains fit the new environment; firms that rely on protection alone are exposed to the next change of policy.

14. Recommendations, Capital

These are themes for research and questions for your advisers, not investment advice.

Immediate (within 6 months). Separate exposure to companies that benefit from strategic-openness policy because they are productive from those that benefit only because they are protected. The first group compounds; the second is a bet on policy continuity with an uncertain end date. Review holdings for single-country input dependence in strategic sectors.

Build (6–24 months). Research the suppliers of the Productive Citizen Economy: industrial automation, robotics, machine vision, industrial software, power electronics, grid equipment, and training and certification providers. Labour scarcity in ageing economies is a structural tailwind for their order books. Look at companies that help migrants reach employment faster, from qualification recognition to language and job-matching services, where demand is set by demographics rather than cycles.

Position (2–10 years). Expect periodic write-downs and cancellations among heavily subsidised strategic projects; avoid concentrated exposure to any single flagship. Prefer firms that sit inside supplier networks in Germany, Japan, Korea and Central Europe over standalone ventures. Treat the EU's 20% manufacturing target as aspirational; companies whose valuation assumes it will be met carry policy risk.

Avoid. Assuming protection guarantees returns. Treating cheap-labour manufacturing locations as automatically lower risk: trade policy, export controls and logistics now carry costs that were absent a decade ago.

Why. Capital earns most where it multiplies scarce people. In ageing economies, that is the combination of automation, skilled labour and secure supply chains, not protection as such.

15. What Would Change Our Mind

Automated reshoring keeps failing. If, over the next five years, most highly automated plants opened in Europe, Japan or Korea to serve domestic markets close or relocate within a few years, as the Speedfactory did, the case for domestic production in strategic ecosystems would weaken, and specialisation through trade would look like the better path.

Automation fails to reach wages. If the labour share of income falls substantially in Korea, Germany and Japan while robot density rises, the claim that labour scarcity steers automation toward multiplying workers would be wrong, and the Stress scenario would become the base case.

Low-skilled migration proves to raise native productivity broadly. Strong evidence that large-scale, less selective migration raises output per worker and wages for natives in ageing economies would weaken the emphasis on selection.

Strategic subsidies show high success rates. If most large subsidised semiconductor, battery and green-steel projects started in 2022–2026 are operating at planned capacity by 2030, the execution risk we emphasise would be overstated, and the probability of the Base scenario would rise.

16. Bottom Line

The old trade-off weighed cheaper goods against lost jobs and assumed the jobs would come back in another form. In an ageing economy they do not, because the young workers who would have filled the new jobs are not there, so each good job that leaves takes with it a taxpayer, a pension contributor and a customer who will not be replaced. That makes a deliberate middle path rational: buy ordinary goods wherever they are made best, keep a short list of strategic capabilities at home or among partners, put robots and AI behind fewer, better-paid people, and select and convert migrants quickly into self-sufficient residents. The direction is right; the record of governments picking projects is not good, so the test of this policy is not how many plants it subsidises but whether output per worker, real wages and the number of productive residents rise together.

Forecast record

Forecasts entered in this article, in resolvable form. Each has a threshold, a named verifier and a resolution date, so it can be scored.

Correlation note: F1 is a policy-process forecast and largely independent of the others. F2 and F4 share a driver, European industrial investment, and are mildly positively correlated; a strong investment cycle would raise both. F3 depends on Japan's labour market and migration policy and is independent of F1, F2 and F4. Entry into the public Ledger is a founder decision.

Sources and epistemic class

Every figure carries its class: measured (reported by the statistical body), projected (an official body's projection), reported (secondary source reporting primary data or analysis), derived (our calculation from cited figures), analytical (argument, not data).

Series I · The Shrinking Core

1. The Wrong Denominator — why headcount no longer measures an economy

2. Globalise Goods, Select People, Protect Capabilities — the Productive Citizen Economy and its limits (this article)

3. Same Machines, Different Futures — how AI will split twelve economies

4. 2030, 2035, 2040 — forecasts and a playbook for the shrinking Economic Core

THRIVE IN CHAOS  ·  Analysis → Forecast → Recommendations  ·  Signal Over Noise  ·  thriveinchaos.ai

AI intelligence system with human editorial oversight. This article is analysis, not investment, legal or financial advice.