Private Debt and Weaker Bargaining Power Shifted Growth Away From Workers
Economist Steve Keen argues that today’s economic strain reflects decades of weaker worker bargaining power, rising private debt and the loss of productive capacity—not simply higher prices or changing expectations. He warns that AI could deepen the problem if it displaces workers without giving them another way to share in the income it generates. Climate risks and war-related supply disruptions, he says, expose how vulnerable economies are when production depends on fragile systems and concentrated sources of essential inputs.

Debt and weaker bargaining power changed who benefits from growth
Steve Keen argues that the economic strain younger people face is not simply a matter of expensive housing or changing expectations. He connects it to a longer shift in economic policy and in the distribution of income: growth slowed, workers lost bargaining power, and more of their income became committed to debt and costs that had previously been publicly provided.
Keen does not claim that young people today have worse lives than workers 150 years ago, when child labor and dangerous mining were common. His comparison is with the expectations his own generation grew up with. When he was 18 and studying at university in Australia, he recalls, unemployment was about 1.5%. He felt almost certain he would find a job and did not worry that a decision he made would jeopardize that possibility.
Since the 1970s, Keen says, economic policy has prioritized controlling inflation over keeping unemployment low. The political promise behind the shift was that deregulation and privatization would produce faster growth and higher wages, making public provision less necessary. People would be able to save for themselves, pay for private hospitals and buy education. Keen says that promise failed. He estimates that per-capita income growth fell from roughly 3% a year between 1945 and 1975 to about 1% a year afterward. He cautions that he is speaking from memory about the precise figures, but says the scale of the change is clear.
The loss of security, in his account, is not just about the availability of jobs. Workers also lost the collective bargaining power that had allowed them to claim a share of economic growth. As unions weakened and gig work spread, workers increasingly faced large employers as individuals, with less leverage to negotiate wages or stable conditions. The promise was that a less regulated economy would deliver rising wages and security; Keen says the experience has instead been greater insecurity and weaker bargaining power.
His family’s finances make the generational contrast concrete. Keen’s father was a bank clerk and the sole earner after his mother left the bank when she married, under the rules of the time. On one income, his father supported a family of seven and bought a house. Keen notes that his father had an advantageous interest rate as a bank employee. His father retired at 55 after becoming ill. The broader arrangement—a single income supporting a large family and a home—is what Keen sees as largely out of reach now. Younger households, he says, often need two incomes, borrow for education and housing, and carry much more debt.
Housing is one place where Keen sees the change in the arithmetic of entry. He recalls a rented share house he and friends could have bought in his early twenties for about $12,500, around one year of his income at the time. He estimates that buying the same house today would require saving the equivalent of ten years’ income. Rising prices can continue, he argues, only while buyers can borrow more than previous owners did and while a new cohort can afford to enter the market. He sees both conditions being strained: the share of people who can buy is shrinking, while the number of people old enough to consider a mortgage is growing.
That is why Keen sees the model of indefinitely rising house prices reaching its limits. The same process that rewards existing owners makes entry harder for people who have not accumulated property or borrowing capacity. Keen says most people are already excluded from the housing ladder; continued price increases require a succession of buyers able to take on still more debt.
He also disputes the idea that the strain on families can be explained mainly by a refusal to accept a lower standard of living after having children. Williamson suggests that some households try to preserve pre-children spending on holidays, restaurants, cars and clothes rather than redirecting it toward family life. Keen agrees that expectations matter, but says families now also have to cover costs that were more often publicly provided in the 1950s and 1960s, including education and health care. Expenses that did not previously come out of a household’s wages increasingly do so now.
That shift would have been easier to absorb if wages had kept pace with economic growth. Keen says real wages and GDP began to diverge around 1974. Workers’ incomes rose somewhat, but he argues that average working-class people are not substantially better off in real terms than they were 50 years ago. His point is not that families should never cut back after having children. It is that wages expected to cover family life have not risen in line with the economy, while more essential costs have been transferred to households.
Keen’s account of why workers lost ground is tied to his research on private debt. He says he began by studying financial instability, adding banks to a cyclical economic model in which businesses borrowed to build factories. The borrowing expanded productive capacity; it was not, in the model, being used to buy yachts or other consumption goods.
Keen expected that as private debt rose, the additional interest going to banks would reduce the income left for capitalists. When he simulated the model, he says, bankers’ income did rise, but the group whose share fell was workers. His explanation rests on bargaining power and the way income is divided. Businesses can pursue their desired level of investment and profit without caring much about how the rest of national income is allocated. But rising debt means more income has to be paid out as interest. If workers cannot bargain to protect their share, the growing claim of creditors comes partly at their expense.
Keen connects this to Marx’s emphasis on investment as a force shaping both income distribution and total income. His argument is not simply that capitalists take income from workers. When borrowing finances investment and debt accumulates, bankers and other rentiers receive a growing stream of interest payments. Firms may still invest at their desired rate of profit, while workers—who, in Keen’s account, have less power to negotiate—absorb the reduction in the share left over.
He therefore argues that creditors can be central to the distributional conflict even when borrowed money is being used to expand production. The factories in his model were productive assets. The mechanism he describes is the increased claim on income created by servicing a larger stock of private debt. Keen says the process shifts income from workers toward bankers and rentiers; it is one reason he rejects the idea that rising economic output will automatically make workers better off.
The UK’s household squeeze reflects a loss of productive capacity
Keen singles out the UK as an especially stark example of a country where living standards have not kept pace with wages and costs. He describes moving from a professorship in Sydney to Kingston University in 2014. His Australian salary was A$170,000, and Kingston agreed to match it with £88,000. After deductions, he says, he took home about £4,000 a month. Rent and other costs left him with roughly £2,000 to live on. Even as the highest-paid academic at Kingston, he says, he could afford to eat at a restaurant once or twice a month, compared with nearly every day in Sydney. Two economist friends who had moved from Australia made a similar estimate: they felt about 40% worse off in London.
Keen’s broader explanation is that the country has de-industrialized and become increasingly financialized. In his view, manufacturing once supported productivity and rising wages; the shift toward services did not replace that productive capacity.
He challenges the idea that manufacturing could simply be moved abroad while the UK retained design and other higher-value services. Design itself depends on industrial capabilities and engineering expertise, he argues, and the West, including the UK, has lost some of the engineers needed to do that work. He sees financial services as a particularly poor substitute: what counted as a major growth engine in the UK, he says, was in large part the creation of private debt. The expansion of finance did not compensate for the decline in manufacturing.
Keen points to aircraft production as an emblem of the change. As a child, he watched a drama about an English aircraft manufacturer and says it would once have seemed normal for the UK to make planes. He contrasts that with the country’s current role, which he describes as producing components for Airbus. His point is not that every part of manufacturing can or should remain in one country. It is that the belief that services could expand indefinitely while industrial capacity moved elsewhere was a mistaken economic strategy.
That structural account also informs his response to Williamson’s example of a small business facing higher wage costs. Williamson describes a pie-shop owner who says that a minimum-wage increase raises staff costs, which may feed through suppliers and prices, and ultimately require the shop to sell many more pies just to maintain its previous position. Keen’s wider argument is that the UK’s problems cannot be reduced to the decisions of workers or small employers. The productive framework in which they operate has been weakened over decades.
The same preference for private provision, Keen argues, has affected long-lived infrastructure. Private managers may have a stronger incentive to produce returns during their tenure than to fund maintenance whose benefits will arrive years later. He points to UK water infrastructure, claiming that after roughly 40 years of privatized water the country has one fewer dam than before, despite having 12 million more people. He also says sewage pollution in rivers reflects a failure to maintain infrastructure over the long term.
Keen does not say public institutions never waste money. He recalls seeing poor performance in a government department where he worked. His objection is to the blanket claim that private provision is always more efficient. For infrastructure that needs maintenance over decades, he argues, the relevant question is whether the organization has an incentive to leave a functioning system to its successor. In his account, the UK’s industrial and infrastructure problems come from choices about what to build, what to maintain and which forms of return to prioritize—not simply from individuals failing to economize.
AI could expand production while weakening the incomes that sustain demand
Keen calls AI an epochal technology, but distinguishes its productive potential from two economic risks: an investment boom that may end in a bust, and automation that could weaken the wages sustaining a mass-market economy.
He sees evidence of real capability in software. A computer programmer he knows uses AI agents to search for problems in how software integrates with other systems, and the tools are now producing code the programmer considers decent. Keen says this can replace work that would otherwise require people. He is more cautious about robotics: he points to unresolved issues with joints, wear, hand movement and maneuverability, and does not claim that machines are already close to replacing all physical labor.
If those problems are solved, however, machines could take over both manual production and clerical processes inside companies and government. Workers have historically been employed because they could do things machines could not. If machines could perform those tasks, that reason for employing people would weaken. The economic difficulty would not be that machines failed to produce goods. It would be that people without jobs would lose wages and therefore the ability to buy them.
Keen describes this as a demand-side problem. Large retailers exist in a mass-market economy because working people can afford goods and services. If automation removes jobs without replacing the income attached to them, businesses may be able to produce more while losing customers. The central economic challenge, in his view, is not only how AI changes production, but who owns the systems and receives the income they generate.
That concern is distinct from his prediction about the AI investment cycle. Drawing on Joseph Schumpeter, Keen describes a new technology as a source of investment-led booms. Entrepreneurs with ideas but little money borrow to put them into practice. Bank lending creates new money, which is then spent on construction, equipment and labor. The railway boom of the nineteenth century is one example he uses; data centers, GPUs and wiring are contemporary counterparts in his account.
The boom does not require every entrant to be foolish. In an uncoordinated market, Keen says, many firms may independently decide to enter an industry that can sustain only a few. Each expects to be among the survivors. As investment accumulates, the cost of building capacity can outrun the income it generates. Keen says he has seen estimates suggesting current AI investment costs are around five times its income, but does not track the figures closely enough to offer a precise estimate.
When the technology comes online, it can also displace established businesses. Keen’s railway example includes canals and carriages, which lost business as rail transport expanded. Investment may be excessive relative to the demand investors anticipated even as the technology itself proves transformative. Keen says some AI firms could fail and share prices based on expectations of sustained growth could fall sharply. He does not say that an investment bust would prove AI useless.
He estimates that data-center investment may amount to 2% to 3% of GDP. If investment fell from 2% of GDP to half a percent, he says, the reduction in demand could contribute to a serious downturn. He compares this with the telecommunications boom of the 1990s, when companies laid far more optical cable than was needed. Much of it had no lasers attached and was called “dark fiber.” After the slump, the excess capacity could be bought cheaply and made usable.
That example explains why Keen separates a financial bust from the longer life of a technology. The boom can leave behind infrastructure that becomes useful later, once it is cheaper. In his reading of Schumpeter, technology spreads through society during the downturn that follows the investment boom, not only while investors are pouring money into it.
Keen links this pattern to Hyman Minsky’s account of financial instability. New opportunities send signals that encourage investment, while finance makes it possible to act on those signals. The resulting boom encourages further investment, including investment beyond what the market can support. Keen argues that mainstream economics fails to understand this instability because it assumes economies tend toward equilibrium.
The employment problem remains even if some AI businesses fail and adoption proceeds more slowly. Keen expects AI to amplify a longer decline in bargaining power for people who do not own capital. If automation reduces the need for labor, the distribution of income becomes more consequential: people still need purchasing power even if their work is no longer required to produce the goods they consume.
A basic income becomes necessary if automation displaces most workers
Keen distinguishes a guaranteed basic income from a universal basic income. A guaranteed basic income, in his usage, would provide support to people who lose their jobs so they can remain able to meet basic needs. A universal payment would go to everyone, including the wealthy. He says he would prefer a guaranteed income under current conditions, and worries that an untargeted universal payment could simply push up prices.
If automation eventually left most people unable to find work, Keen says, the case would change. A universal income would then be necessary to provide purchasing power to people displaced by technology. Otherwise, firms might produce goods for a small group of owners and workers while a much larger population could not afford to buy them. The state, in his view, would have to create money people could spend.
That is why he criticizes Elon Musk’s support for a high-income future alongside efforts to cut government spending. Public money creation would be needed to make a society in which most people did not work economically viable, Keen says. Williamson asks whether the goal of DOGE was not to improve efficiency. Keen replies that efficiency efforts should be led by engineers, and objects to the people he says Musk put in charge. He also says that government waste exists, but is not the central problem.
Keen disputes the analogy between governments and households. He says governments create money rather than borrowing it in the same way a household or firm does. This does not mean that any level or form of spending is automatically useful. His distinction is between waste within a public agency and the broader claim that government spending is inherently unsustainable or that private firms are always better suited to provide public goods.
The infrastructure question connects the argument about income support to Keen’s criticism of privatization. If automation concentrates ownership and income, public provision may be needed to sustain demand. If long-lived systems require maintenance over decades, he argues, public institutions may also be better placed to maintain them than managers whose returns are judged over shorter periods. Both claims depend on how institutions are organized and what incentives they face, rather than on a simple opposition between efficient markets and wasteful government.
Keen presents universal income as conditional, not as an immediate solution to every economic problem. It becomes necessary in the scenario where technology takes over most work and ordinary wages no longer distribute enough income for people to participate in the economy. His criticism is that political efforts to weaken government could remove one of the tools needed to manage that transition.
Climate risk is not simply a small adjustment to economic growth
Keen’s climate warning rests on a criticism of how economists estimate economic damage. He says economists have measured differences in income between places with different temperatures and used those relationships to estimate the effect of future warming. He objects to treating what happens “across space” as a forecast of what will happen “through time.” A modest relationship between temperature and income across locations today, he argues, cannot establish how economies will respond when climate systems change over time.
Keen cites an estimate in the 2022 IPCC report that four degrees of warming by 2100 would reduce income in 2100 by between 10% and 23%. He argues that the figure can sound large while implying a small reduction in the annual growth rate over the decades before 2100. Using an illustrative income of 10 growing toward 100, he says, an outcome of 90 rather than 100 at the end of the period produces only a small difference in annual growth. He describes the implied reduction as about 0.05 percentage points a year.
He contrasts that economic framing with warnings from climate scientists. Keen says some scientists describe 1.5 degrees of warming as dangerous, three degrees as catastrophic, and five degrees or more as carrying existential risk. He also refers to a 2021 survey of climate economists: respondents considering a scenario of seven degrees of warming by 2220 expected GDP to be about 25% smaller. Keen argues that translating such a loss into a small change in growth over two centuries obscures the possibility that warming could disrupt the physical systems on which economies depend.
One and a half degrees of warming is dangerous, three degrees is catastrophic, and five degrees and above is beyond catastrophic, including existential risks.
When Williamson asks what could make him wrong, Keen answers that climate scientists might not understand the climate. He then describes the development of climate modeling as a response to the nonlinear feedbacks in weather systems. Before the 1960s, he says, weather forecasting relied on pattern matching: a sequence of conditions in the past was used to infer what might happen next. Meteorologist Edward Lorenz instead modeled weather as a nonlinear system. Keen describes Lorenz reducing a more complex set of fluid-dynamics equations to a simple three-equation model and finding chaotic behavior that surprised him. Keen says that work helped establish the modeling approach used in weather forecasting and climate prediction.
The consequence Keen focuses on is the Atlantic Meridional Overturning Circulation, or AMOC. He describes it as a large-scale ocean circulation beneath the surface flows more commonly associated with the Gulf Stream. Warm, salty water moves north, transfers heat to the atmosphere, cools and becomes dense enough to sink, then circulates back at depth. Keen says the system carries a volume of water many times greater than the combined flow of the world’s rivers and has stopped at points in the planet’s past.
Keen cites research suggesting that the AMOC could shut down sometime between 2025 and 2095, with 2050 as a best estimate. He says a shutdown could lower Europe’s temperature by 10 degrees Celsius over three decades. He describes the resulting range of average temperatures in London as between minus 20 and plus 25 degrees, and says London and Europe would be uninhabitable under that scenario.
He further cites work by Tim Lenton, a climate scientist at Exeter, concerning a combination of AMOC shutdown and 2.5 degrees of warming. Keen says Lenton estimated that the area of the planet able to support grain could fall by 70%, with global famine as the consequence. The point Keen draws is that estimates implying a minor slowdown in growth do not capture the consequences of abrupt physical changes to climate systems and food production.
Keen argues that policymakers rely too heavily on economic estimates that make climate change appear distant and manageable. He says scientists he knows are discussing risks on a 10-to-25-year horizon, rather than only in the far future, and that politicians can respond to reassuring economic projections while continuing to approve fossil-fuel projects. Williamson mentions a previous conversation with climate scientist Richard Betts, who did not give a definitive forecast of future temperatures or carbon-dioxide levels. Keen replies that climate science continues to develop and that the scientists he knows are increasingly alarmed.
His argument is about the limits of the economic frame as much as a forecast of a particular climate outcome. A calculation of lost income does not, by itself, describe the risk of changes to ocean circulation, agriculture or other systems. Keen sees climate damage as a potential constraint on the ability to produce and distribute goods, not merely as a percentage reduction in future GDP.
War threatens the inputs that keep food and industry moving
Keen describes the war involving Iran as a strategic mistake and a possible shock to supply chains well beyond oil. He says the country’s mountainous geography, population of about 90 million and industrial capabilities make it unlike Iraq, which he says could be invaded and its government rapidly overthrown. He also asserts that Iran had prepared for the possibility of its leadership being killed by dividing its military into 31 regional units, each associated with one of the country’s regions, so they could act if central command was decapitated. In Keen’s account, that preparation makes decentralized resistance more likely than a swift collapse.
He challenges Western assumptions about Iranian society, saying women make up almost half of its STEM graduates. His broader claim is that Iran has substantial technical and industrial capacity and is not a small or easily occupied state. He calls the decision to attack reckless.
When Williamson asks why the war began, Keen speculates about political pressure, Israel’s strategic interests and possible blackmail. He says he cannot make sense of Western leaders’ decisions and suggests several possible motives. He argues that Israel seeks regional dominance and says its expansionism helps explain its hostility toward Iran. He also says public views of Israel have changed as people see images from Gaza, while politicians have criticized people who support Palestinians.
Keen’s most concrete near-term warning concerns fertilizer. He says a significant share of fertilizer is produced in the Middle East and that production depends on oil-derived feedstocks and sulfur. The region exports fertilizer, particularly to Southeast Asia. If transport is blocked and production is disrupted, fertilizer becomes both scarcer and more expensive.
Keen says colleagues in agriculture have told him that, without artificial fertilizer, the planet’s carrying capacity under conventional pre-fertilizer agriculture would be one or two billion people, compared with a current population of eight billion. He estimates that the war has cut off 20% to 30% of global fertilizer production and predicts that shortages could lead to famine, with effects beginning in 2027.
He emphasizes that the impacts would arrive with a lag. Crops already planted may not show the effect immediately, but disruption to fertilizer inputs can affect harvests months later and carry forward into subsequent growing seasons. A crop’s growing cycle means that a supply shock can be underway before its consequences become visible in food output.
The supply-chain concerns extend to helium and semiconductor production. Keen says helium is brought to the surface with oil, is difficult to separate, and is used to cool equipment in semiconductor manufacturing to temperatures just above absolute zero. He says attacks have damaged some helium-extraction equipment and that supplies are not moving through the straits. As helium is lost through leakage in manufacturing, shortages could eventually limit chip production, including the GPUs used for AI.
He also names sulfuric acid, used in manufacturing processes including copper refining, and lubricating oils produced around the Gulf. If stocks run down, he says, manufacturing equipment may stop working even when factories are intact. His point is that “oil” is not one interchangeable commodity. Keen says oils differ by source, and that the mix produced in the Middle East has properties that fracking, Venezuelan and Canadian oil do not supply in the same way.
Fuel availability is already a concern in his account. Keen singles out diesel as essential for farm machinery and shipping. Less diesel would mean less ability to grow food and move it, not only higher prices at the pump. He describes a possible 20% to 30% fall in fuel availability as an unusually serious disruption outside wartime, then questions whether the situation should be called peacetime at all.
He invokes Dwight Eisenhower’s warning about the military-industrial complex and argues that war spending benefits weapons manufacturers while working-class people bear the costs of fighting. In Keen’s account, the risk is not limited to higher prices: production depends on specific inputs, transport routes, equipment and inventories. A disruption can reduce what can be made or moved, even before its full effect is visible in consumer prices.
Work is also how people are taught to measure their worth
The prospect of work becoming optional raises a question Keen treats as social as much as economic: would people flourish without jobs, or lose a central source of meaning? He argues that the problem is already visible in a society where people who do not work are often treated as if they have nothing to contribute.
Keen looks to hunter-gatherer societies as a contrast. He describes small communities in which people gained status by contributing useful work to one another. A skilled flint maker might make tools for skilled fishers and receive fish in return. In the societies he discusses, work was embedded in reciprocal relationships and community respect rather than organized around a paid job. He says people who contributed nothing risked being excluded, but contribution could take different forms, and the community was much more tightly connected.
For Keen, the point is not that everyone would become an entrepreneur or artist if paid work disappeared. He rejects the idea that everyone wants to build a business. Some people might create, others might watch sports or take part as spectators. A society with less necessary labor would still need ways for people to participate and be valued.
That distinction informs his claim that economic value and personal worth have to be separated. He and Chris Williamson agree that “What do you do for a living?” makes a job the default measure of a person. Williamson offers “What keeps you busy?” as a better question: it allows someone to describe work but also interests, routines and commitments outside employment. Keen says that is closer to what human life should be about.
Keen imagines a future in which machines perform necessary manual labor and people have more scope to be thinkers, artists or other kinds of participants. But he worries that this outcome would not arise automatically. Without a shared culture to guide life beyond work, abundance could become bedlam rather than freedom.
His proposed response is “Saganism,” a new civic religion built around science, knowledge and the preservation of life. Keen argues that science currently lacks the cultural presence religion has had for most of human history. Williamson pushes back that religion often works through story, myth, archetype and personification—forms that may engage people more readily than scientific explanation. Keen agrees that science needs to become more compelling, but rejects the idea that it must imitate supernatural claims.
He suggests treating experiments as science’s counterpart to religious miracles: an experiment can be repeated and tested. Scientists who challenge accepted views, he says, face ridicule and abuse and should be regarded as exemplary figures. He proposes Carl Sagan as one possible symbolic figure and the image of Earth as a “pale blue dot” as a shared emblem. The image, in Keen’s account, holds two ideas together: everything people have loved and fought over has happened on one small planet, and that planet is the only place in the universe where life is known to exist. A science-centered culture, he hopes, would make protecting life and exploring the universe a common purpose.
Williamson raises the risk that calling science a religion could make it seem sterile. Keen’s answer is that science should be made emotionally and culturally compelling rather than reduced to technical facts. His proposal is not a detailed institutional plan, but a response to the possibility that automation could weaken the role work plays in identity and community.
The next decade could expose how dependent production is on fragile systems
Asked what the next five to ten years may look like, Keen points first to climate effects and the war’s consequences for productive capacity. He predicts that fertilizer and fuel disruptions could contribute to food shortages, though he says which countries are affected will depend partly on who can afford fertilizer. He identifies parts of Southeast Asia as vulnerable to being unable to buy enough. He also warns that the UK could face food restrictions if weather volatility compounds the loss of agricultural inputs.
Weather variability is a separate pressure in his account. Williamson notes that the UK had a period in 2026 when it rained every day until March. Keen says conditions then became dry and adds that an El Niño may bring heavy rainfall. He argues that volatile rainfall can make it harder to produce food even before fertilizer shortages are considered. He says the combination of war-related supply disruption and climate volatility could mean mass famine in some places and restrictions on food consumption in others.
Keen also refers to public advice to keep a few days of food at home as evidence that governments are considering the possibility of shortages. The timing of any disruption matters: a crop’s growing season creates a lag between reduced fertilizer availability and its effect on harvests. Keen says people often mistake delay for an absence of damage.
If the AMOC were to shut down, Keen says, the consequences could be far more severe, potentially making the current economic order impossible to sustain. AI represents a different possibility. He imagines machines doing necessary work and freeing people to pursue other activities, but says the transition would require a way to distribute income and give people a role beyond employment. Without that, automation could intensify the loss of purchasing power and social purpose.
On Bitcoin, Keen’s concern is that its energy use could make it vulnerable in a future where energy consumption has to fall. He says he could have bought Bitcoin when it was worth about a pound but did not, in part because of the energy required by its security mechanism. He estimates its energy use as comparable to Switzerland’s and says roughly 2% of the population holds it. He argues that a currency system that excluded most of the population would not be socially viable.
If societies conclude that energy use has to be cut sharply, Keen says, international travel and Bitcoin could be easy targets. He estimates that restricting those activities could reduce energy consumption by 6% or 7%. His point is not a detailed prediction of a Bitcoin price, but that a system dependent on substantial energy use could face political restrictions in a crisis.



