This paper asks a simple question with complicated answers: when a t-shirt is sewn in Dhaka and sold in Berlin, or a smartphone is assembled in Zhengzhou and sold in Chicago, where is the value produced, and where does it end up? Mainstream research on #global_value_chains describes this system in terms of "value added," "governance," and "upgrading." Marxist political economy describes it in terms of #surplus_value, #exploitation, and #unequal_exchange. The two languages produce very different pictures of the same trade flows. This article develops the second picture. Working from the categories set out in Marx's Capital, and from the dependency tradition associated with Marini, Amin, and Emmanuel, it argues that the geography of value production and the geography of value realisation have been systematically pulled apart. Value is produced disproportionately in low-wage production zones of the #Global_South and recorded disproportionately as income in the #Global_North. The paper identifies eight concrete mechanisms through which this separation is organised: wage suppression below the value of labour power, cross-border #labour_arbitrage through arm's length contracting, #monopsony buyer power and the cost-price squeeze, #intellectual_monopoly and brand rent, financial channels including #transfer_pricing and profit shifting, the compression of circulation time through #logistics, ecological appropriation, and the unpaid labour of #social_reproduction. It illustrates these mechanisms through two sectors that define contemporary capitalism, garments and electronics, and one input sector, battery minerals. It then confronts the strongest objections to the argument: that value transfer cannot be measured, that upgrading is real, that China complicates the core-periphery map, and that wage gaps reflect productivity gaps rather than power. The conclusion is that dependency theory does not need to be abandoned, but it does need to be rebuilt around the firm-level architecture of the modern supply chain rather than around the nation-state as the unit of analysis. Keywords: global value chains, surplus value, unequal exchange, dependency theory, super-exploitation, labour arbitrage, imperialism, fast fashion, electronics manufacturing, monopsony 1. Introduction There is a familiar number that circulates in trade policy discussions: roughly seventy percent of world trade now consists of intermediate goods, services, and components rather than finished products moving from producer to consumer. Production has been sliced into tasks, and those tasks have been distributed across borders according to cost. This is normally described as a story about efficiency. Firms locate each stage of production wherever it can be done most cheaply, and everyone gains something. This paper takes that same reality and reads it through a different set of categories. The claim is not that #global_value_chains are inefficient. The claim is that they are a machine for moving #surplus_value across borders, and that the direction of movement is not random. It runs from places where labour is cheap, disciplined, and often informal, toward places where capital is concentrated, brands are owned, patents are held, and profits are booked. That movement is what the older literature called #unequal_exchange, and the argument of this paper is that it has become more important, not less, since the term fell out of academic fashion in the 1980s. The stakes are not merely definitional. If value is created where it is recorded, then the enormous income gap between a design office in Portland and a sewing floor in Chattogram simply reflects a gap in productivity, and the policy conclusion is to make the sewing floor more productive. If value is created in one place and recorded in another, then the same gap reflects a transfer, and the policy conclusion is entirely different. The measurement convention is doing political work. This paper argues that the convention is wrong, and that the wrongness is not accidental. 1.1 The problem stated plainly Consider two workers. The first assembles smartphone components on a production line in an industrial park in central China, working long shifts with compulsory overtime, living in a company dormitory, earning a wage that covers food, rent, and remittances home but little else. The second is a software engineer at the firm whose logo appears on the phone, working in California, earning perhaps forty times as much per hour. Standard national accounts will record the Californian as vastly more productive. But productivity here is measured in money, and money value is assigned by the accounting system rather than discovered by it. The assembly worker is running a line that produces hundreds of units per shift. The value of those units, once realised in a retail store, is recorded as revenue in the accounts of the brand-owning firm, not the contract manufacturer, and certainly not the worker. What appears in the statistics as high Northern productivity may be, at least in part, Southern #surplus_value that has been relabelled during its journey through the chain. This is the intuition. The rest of the paper tries to turn it into an argument, with mechanisms, evidence, and a serious treatment of the objections. 1.2 Contribution and structure The paper makes three contributions. First, it synthesises a literature that is currently fragmented across at least four separate conversations: quantitative work on ecological and monetary #unequal_exchange, labour studies research on purchasing practices in supply chains, political economy of #intellectual_monopoly, and the revived dependency research programme. These conversations rarely cite each other systematically. Second, it specifies mechanisms rather than asserting outcomes. Much writing in this area states that value is transferred without explaining how, which leaves the argument vulnerable. Third, it takes the counterarguments seriously and concedes what should be conceded. Section 2 reviews the literature. Section 3 sets out the theoretical framework drawn from Capital. Section 4 explains the method. Section 5, the core of the paper, details eight mechanisms of #value_transfer. Section 6 works through sectoral cases. Section 7 examines why standard statistics conceal the process. Section 8 discusses what all this means for dependency theory and confronts objections. Section 9 turns to political and policy implications. Section 10 states limitations. Section 11 concludes. 2. Literature Review 2.1 The mainstream global value chain and production network literatures The dominant framework for studying cross-border production emerged in the 1990s and 2000s and matured into the #global_value_chains approach, with a parallel and partly overlapping global production networks tradition. Its central concepts are governance, the way lead firms coordinate suppliers without owning them, and #upgrading, the idea that suppliers can move into higher-value activities over time. The framework has been enormously productive empirically. It gave researchers a vocabulary for describing how a firm in Amsterdam can control production in Vietnam without holding a single share in the Vietnamese factory. But the framework carries an implicit theory of value that is rarely defended. It treats "value added" as if it were value created, and it treats the distribution of value added across the chain as a matter of bargaining position, market structure, and capabilities. In this account, if a supplier captures only a small share of the final price, that is because the supplier performs low-skill tasks that are inherently worth little. The measure and the explanation are the same thing, which makes the argument circular. Critical scholars working within and against this tradition have made the point forcefully. Baglioni, Campling, and Hanlon (2020) argue that lead-firm dominance is better understood as entrepreneurial capture, an active appropriation of value produced elsewhere rather than a passive reflection of where value originates. Selwyn (2019) coined the phrase "poverty chains" to describe the same chains from the standpoint of the workers inside them, noting that integration into world markets has repeatedly coincided with wage suppression rather than wage convergence. Selwyn and Leyden (2022) show that the development orthodoxy which recommends deeper #global_value_chains integration systematically underplays the oligopolistic structure of lead-firm markets. Quentin and Campling extended the analysis toward what they called global inequality chains, drawing the fiscal and tax dimensions into the picture. 2.2 The Marxist and dependency traditions The older tradition asked different questions. Emmanuel's theory of #unequal_exchange argued that when profit rates tend to equalise internationally but wage rates do not, commodities produced in low-wage countries exchange at prices below their value, and commodities produced in high-wage countries exchange above theirs. The result is a systematic transfer embedded in ordinary trade at market prices. No cheating is required. Amin generalised this into a theory of imperialist rent grounded in monopoly, arguing that the barrier to entry protecting Northern firms is what allows the transfer to persist. Marini, writing in Latin America, developed the concept of #super_exploitation, the payment of labour power below its value, as the specific mechanism by which dependent economies compensate for the value they lose in trade. His work has recently become far more widely available in English translation (Marini, 2022), which has helped drive the revival discussed below. These theories were criticised heavily. Some critiques were technical, concerning the treatment of values and prices. Others were political, arguing that the framework encouraged national developmentalism and obscured class conflict inside the periphery. By the 1990s dependency theory was widely treated as a dead research programme. That judgement has been reversed. Kvangraven (2021) argues persuasively that the caricature of dependency theory as a crude, stagnationist, nation-versus-nation doctrine bears little resemblance to the actual research programme, which was internally diverse and centrally concerned with the structural constraints that global capitalism imposes on peripheral accumulation. Madariaga and Palestini (2021) show the framework's continued analytical purchase in both Latin America and Europe. Patnaik and Patnaik (2021) reconstruct the theory of #imperialism around the role of the periphery in disciplining the price level of the core. 2.3 The quantitative revival The most striking development in the last few years is the return of measurement. Hickel, Sullivan, and Zoomkawala (2021) estimated the scale of drain from the #Global_South through #unequal_exchange between 1960 and 2018, finding that in 2017 alone the South lost resources worth roughly 2.2 trillion United States dollars when valued at Northern prices, and that the cumulative drain over the period reached tens of trillions. Hickel, Dorninger, Wieland, and Suwandi (2022) extended the analysis to physical flows, finding that in 2015 the North net-appropriated from the South billions of tonnes of embodied raw materials, hundreds of millions of hectares of embodied land, exajoules of embodied energy, and, most importantly for this paper, on the order of 188 million person-years of embodied labour. Dorninger et al. (2021) documented the same asymmetry in ecological terms across a wide sample of countries. Ricci (2019, 2021) approached the problem from a different direction, using input-output data to estimate monetary #value_transfer between countries and finding substantial net transfers toward high-income economies. Carchedi and Roberts (2021) offered a Marxian value-theoretic account and estimate. Suwandi (2019) provided the most detailed firm-level reconstruction, showing how the "labour-value commodity chain" is organised around unit labour cost differentials rather than technology transfer. 2.4 The gap this paper addresses What is missing is a synthesis that connects mechanism to measurement. The quantitative literature demonstrates a pattern. The labour studies literature explains contract-level dynamics. The intellectual monopoly literature explains rent extraction. Nobody has assembled these into a single account of how the modern chain works as a value-transfer apparatus, tested against the sectors where the process is most visible. That is the gap this paper tries to fill. 3. Theoretical Framework: Reading Capital into the Supply Chain 3.1 Value, labour, and the source of profit The starting point is the #labour_theory_of_value. In Marx's account, the value of a commodity is determined by the socially necessary labour time required to produce it under average conditions of productivity and intensity. Labour power, the capacity to work, is itself a commodity, and its value is determined by the labour time needed to reproduce the worker and the worker's family at a historically and socially determined standard of living. The key move is the distinction between labour and labour power. The capitalist buys the second and uses the first. If the worker's wage represents four hours of socially necessary labour, but the worker works ten, then six hours are unpaid. That unpaid labour is #surplus_value, and it is the source of profit, interest, and rent. The ratio of unpaid to paid labour is the rate of exploitation. This is not a moral claim in the first instance. It is an accounting claim about where the surplus in a capitalist economy comes from. The moral weight comes later. 3.2 Absolute and relative surplus value Marx distinguished two ways to increase #surplus_value. Absolute surplus value is increased by lengthening the working day or intensifying work within it. Relative surplus value is increased by raising productivity in the industries that produce workers' consumption goods, thereby reducing the value of labour power and increasing the unpaid portion without lengthening the day. The conventional story is that mature capitalism relies on relative surplus value: machines, not whips. What the evidence from contemporary supply chains suggests is that the two strategies have been geographically separated. The high-productivity, relative-surplus-value regime characterises certain protected activities in the core. The absolute-surplus-value regime, long hours, high intensity, weak or absent labour protections, characterises much manufacturing in the periphery. Both feed into the same chain, and the same firm's income statement. 3.3 The transformation of values into prices of production Volume Three of Capital introduces the mechanism that makes #unequal_exchange possible. Capital moves toward higher profit rates. This movement tends to equalise the rate of profit across sectors. But sectors differ in their organic composition of capital, the ratio of investment in machinery and materials to investment in labour power. Since only living labour produces #surplus_value, labour-intensive sectors produce more surplus value per unit of capital than capital-intensive ones. Equalisation of the profit rate therefore requires that some of the surplus value produced in labour-intensive sectors be realised as profit in capital-intensive ones. Commodities sell not at their values but at #prices_of_production, which systematically redistribute surplus value between sectors. This is the theoretical hinge. Once we accept that the market redistributes surplus value between sectors within a country, there is no principled reason to deny that it redistributes surplus value between countries when capital is mobile across borders and labour is not. And that is precisely the configuration of the contemporary world economy: near-total capital mobility, tightly policed labour immobility. 3.4 Foreign trade as a counteracting tendency Marx himself pointed in this direction. In his discussion of the tendential fall in the #rate_of_profit, he lists foreign trade among the counteracting influences, noting that capital invested in colonial or foreign trade can yield higher rates of profit because it competes with less-developed production and exploits labour paid below the level prevailing in the metropolis, and that the surplus profits so obtained flow back to the metropolitan economy. This is a compact statement of the argument that Emmanuel, Amin, and Marini later expanded. 3.5 Super-exploitation Marini's contribution deserves emphasis because it addresses a specific gap. Standard value theory assumes labour power is bought at its value. Marini argued that in dependent economies, capital compensates for the value lost through trade by paying labour power below its value, through some combination of wage compression below subsistence, extension of the working day, and intensification of labour. This is #super_exploitation. It is not merely a higher rate of exploitation; it is a structural feature that undermines the reproduction of the workforce itself, sustainable only because the wage is subsidised by unpaid #social_reproduction, by rural household production, or by the constant renewal of the labour force through migration. Read alongside the evidence on dormitory labour regimes, informal subcontracting, and homework, this concept describes the modern export factory with uncomfortable accuracy. 3.6 Monopoly, rent, and the modern lead firm Finally, the framework requires a theory of monopoly. Amin's imperialist rent, and the contemporary literature on #intellectual_monopoly (Durand and Milberg, 2020; Rikap, 2021), converge on a common insight. The lead firm's power rests on assets that cannot be replicated: brands, patents, design standards, data, platform position, control of the interface with the final consumer. These assets do not produce value. They allocate it. They function as barriers that allow the firm holding them to command a share of the total surplus value produced across the chain far in excess of the labour performed inside its own walls. This is why the modern lead firm is often asset-light in physical terms and asset-heavy in legal terms. It owns almost no factories and almost all the intangibles. 4. Methodology This is a theoretically driven critical synthesis rather than a primary empirical study. The method has three components. First, conceptual reconstruction. The paper reads the categories of Capital against the institutional architecture of contemporary supply chains, asking in each case what the category implies for the observed structure and what the structure implies for the category. Second, structured synthesis of secondary evidence. The paper draws on three evidence bases: quantitative estimates of monetary and physical #unequal_exchange from the recent input-output literature; firm-level and contract-level evidence from labour studies research on purchasing practices, wages, and working conditions; and corporate financial data as analysed in the literature on #profit_shifting and #intellectual_monopoly. No new data are generated. Where the literature disagrees, the disagreement is reported rather than resolved by assertion. Third, sectoral illustration. Two sectors are examined in depth because they represent the two dominant models of contemporary chain organisation: garments, which is buyer-driven, low-barrier, and highly competitive at the supplier end, and electronics, which is producer-driven at the design end and oligopolistic at the assembly end. A third short case, battery minerals, is included because it shows the mechanisms operating at the extractive base of the chain, where the language of #upgrading has least purchase. The main methodological limitation should be stated at the outset. Value, in the Marxian sense, is not directly observable. It is a social relation, not a price tag. Any estimate of #value_transfer requires assumptions, most commonly about what prices Southern exports would command if wages were equalised, or about the labour content embodied in traded goods. Different assumptions yield different magnitudes. What the estimates can establish is the existence, direction, and rough order of magnitude of the transfer. What they cannot establish is a precise figure. Readers who want precision here will be disappointed, and should be sceptical of anyone who promises it. 5. The Mechanisms of Value Transfer This section is the analytical core. It identifies eight mechanisms. They overlap and reinforce each other, and no single one is sufficient on its own. 5.1 Mechanism one: wage suppression below the value of labour power The most direct mechanism is the wage. Across the major export manufacturing zones, statutory minimum wages sit well below any credible estimate of a living wage, and actual take-home pay often sits below the statutory minimum once deductions, unpaid overtime, and production-target penalties are counted. The point is not simply that Southern wages are lower. Of course they are; prices are lower too. The point is that the gap between wages and any locally calculated subsistence basket is itself large and persistent. When a full-time garment worker cannot afford adequate nutrition, housing, and schooling for dependents on a legal full-time wage, the wage is not paying the value of labour power. The difference is not a market outcome. It is a subsidy, extracted from the worker's own reproduction and from the unpaid labour of the household. This is #super_exploitation in Marini's precise sense, and it is not marginal. It is the normal condition across the export garment sector in South and Southeast Asia, across much of electronics assembly, and across agricultural export production. Sullivan and Hickel (2023) place this in long historical perspective, showing that periods of forced integration into capitalist world markets have repeatedly coincided with declines in real wages and in indicators of human welfare, rather than the smooth improvement assumed by the standard narrative. The mechanism transfers value because the goods produced by these workers are sold in markets where the price is set by conditions in the core. The gap between the price and the wage cost is #surplus_value, and it is realised as profit somewhere along the chain. 5.2 Mechanism two: labour arbitrage through arm's length contracting The second mechanism concerns organisational form. The lead firm does not own the factory. This is the defining innovation of the last four decades, and it is not primarily about efficiency. It is about the legal, fiscal, and moral separation of profit from production. Suwandi (2019) describes this as the labour-value commodity chain: production is relocated not to where productivity is highest but to where unit labour cost is lowest, that is, where the ratio of wage to output is most favourable. The lead firm captures the difference. Because the supplier is a separate legal entity, the lead firm bears no liability for wages, safety, environmental damage, or dismissal. Because there are thousands of potential suppliers and a handful of major buyers, the terms of the contract are set by the buyer. The critical feature is that #outsourcing does not reduce the lead firm's control. Buyers specify materials, tolerances, quality standards, delivery windows, and increasingly the production processes themselves. What the buyer sheds is not control but responsibility. This is what Baglioni, Campling, and Hanlon (2020) mean by capture: the firm exercises the authority of an owner while carrying the liabilities of a customer. 5.3 Mechanism three: monopsony power and the cost-price squeeze Third, and closely related, is the structure of buyer power. In most consumer goods chains, the market has an hourglass shape. Many producers at the bottom, many consumers at the top, and a narrow waist of a few dozen global buyers through which everything must pass. This gives buyers #monopsony power over suppliers. The consequence is what Anner (2020) calls the sourcing squeeze. Buyers demand shorter lead times, smaller minimum orders, higher quality, and social compliance certification, while simultaneously pushing down the price paid per unit. Anner's analysis of purchasing practices in the Bangladeshi garment sector documents a decline in the real price paid to suppliers over recent years, even as compliance requirements after Rana Plaza raised suppliers' costs. Suppliers absorb the squeeze in the only place they can: labour. Wages stagnate, overtime rises, work intensifies, and production is quietly subcontracted to smaller unregistered units where conditions are worse and inspection is absent. The pandemic made the power asymmetry explicit. Anner (2022) documents the mass cancellation of orders, including orders for goods already produced or in production, with costs pushed back onto suppliers and, through them, onto workers who were dismissed without severance. A contract that can be voided at will by one party is not a contract between equals. It is an instrument of #value_capture. The #cost_price_squeeze is therefore not an unfortunate side effect of competition. It is the transmission belt that converts buyer market power into higher rates of exploitation on the factory floor. 5.4 Mechanism four: intellectual monopoly and brand rent The fourth mechanism operates through law rather than through the market. Patents, trademarks, design rights, copyrights, and trade secrets create legally enforced monopolies. Durand and Milberg (2020) show how these intangible assets have become the central axis of value distribution within #global_value_chains, allowing lead firms to protect margins that competition would otherwise erode. Rikap (2021) develops the concept of intellectual monopoly capitalism, showing how a small number of firms have organised innovation systems in which they appropriate knowledge produced across a wide network, including public research institutions and supplier firms, while retaining exclusive legal ownership of the results. For our purposes the key point is analytical. A patent does not create value. It creates a claim on value. When a firm's profits are attributed to its "intangible assets," what is being described in Marxian terms is a rent: a deduction from the total pool of #surplus_value produced by living labour, awarded on the basis of legal exclusion rather than production. The trademark on a running shoe adds nothing to the shoe's usefulness. It adds a great deal to its price, and every unit of that price difference is a claim on surplus value produced by the worker who stitched it. This is why measures of "value added by intangibles" are so misleading. They record the rent and call it production. 5.5 Mechanism five: financial channels, transfer pricing, and profit shifting The fifth mechanism is accounting. Multinational firms trade with themselves. A substantial share of world trade takes place between affiliates of the same parent company, and the prices used in those internal transactions are set by the company. This creates enormous scope for booking costs in high-tax jurisdictions and profits in low-tax ones. Torslov, Wier, and Zucman (2023) estimate that close to forty percent of multinational profits are shifted to tax havens annually. Cobham and Jansky (2020) show that the revenue losses from these flows fall disproportionately, as a share of national income, on lower-income countries. Rikap (2022) shows how platform firms combine intellectual rentiership with this financial architecture. #transfer_pricing is not a separate phenomenon from #unequal_exchange. It is one of its administrative instruments. When a subsidiary in a producing country sells output to a related trading entity in a low-tax jurisdiction at an artificially low price, and that entity resells at market price, the value produced by workers in the producing country is legally relocated before it is ever taxed there. The state in the producing country loses revenue. The workers lose any prospect that the surplus they produced will return to them as public goods. 5.6 Mechanism six: logistics, circulation time, and the compression of turnover The sixth mechanism is less often discussed in value-transfer terms. Marx observed that capital's profitability depends not only on the rate of #surplus_value but on the speed of turnover: how quickly capital advanced returns as money, ready to be advanced again. Circulation time is dead time. Shortening it raises the annual rate of profit without any increase in the rate of exploitation at the point of production. The revolution in #logistics, containerisation, port automation, real-time inventory management, algorithmic demand forecasting, is best understood as a sustained assault on circulation time. Khalili (2020) traces how ports, shipping lanes, and the legal regimes that govern them have been reconstructed to serve this imperative, and how deeply this reconstruction is entangled with state power and military force. The consequence for workers is direct. Compressed lead times mean that demand fluctuations at the retail end are transmitted almost instantaneously to the production end. Buyers hold no inventory risk; suppliers do. Suppliers manage that risk through workforce flexibility, which in practice means short contracts, agency labour, forced overtime during peaks, and layoffs during troughs. The logistics revolution has shifted the entire burden of uncertainty down the chain and onto the least protected people in it. Fast fashion in particular is built on this: the ability to move a design from sketch to shop floor in weeks depends on a supplier base that can be commanded to surge and then abandoned. 5.7 Mechanism seven: ecological unequal exchange The seventh mechanism concerns nature. Dorninger et al. (2021) and Hickel et al. (2022) show that the North is a systematic net importer of embodied land, energy, materials, and labour, and a net exporter of environmental harm. The physical accounting is unambiguous even where the monetary accounting is contested: the goods flowing north embody more resources and more labour than the goods flowing south, and the difference is not compensated by the price. #ecological_unequal_exchange matters for value theory because natural conditions of production are, in Marxian terms, free gifts to capital only in the sense that they are not paid for. Somebody bears the cost. When a river is poisoned by textile dyeing, the cost is borne by the people who drink from it and by the workers whose health is destroyed. That cost never appears in the price of the garment. It is an uncosted subsidy from the periphery to the profit of the core, and it is a form of appropriation as real as the unpaid hour on the sewing line. Fanning and Hickel (2023) extend the logic to atmospheric appropriation, quantifying the disproportionate use of the global carbon budget by high-income economies. 5.8 Mechanism eight: social reproduction and the unpaid labour of the household The eighth mechanism is the one most often missed. Mezzadri (2021) and Mezzadri, Newman, and Stevano (2022) argue that the value produced within global chains rests on a foundation of labour that is not paid at all: the care, cooking, water-carrying, childrearing, and subsistence agriculture, performed overwhelmingly by women, that reproduces the workforce day by day and generation by generation. This is not a supplementary point. It is what makes #super_exploitation possible. A wage below subsistence can only sustain a workforce if subsistence is being topped up from elsewhere. In practice it is topped up by the village, by the extended family, by the woman who works a sixty-hour week in a factory and a further thirty at home, and by informal home-based work that is invisible in every statistical system. The #informal_labour that sits beneath the formal export factory, the homeworkers finishing embroidery, attaching buttons, and hand-stitching in their own rooms at piece rates, is part of the same chain and part of the same value flow. It simply never appears in the audit. 5.9 The mechanisms as a system These eight mechanisms are not alternatives. They are components of a single architecture. Low wages are enforceable because labour cannot migrate and unions are repressed. The wage gap is exploitable because capital can relocate freely. Relocation is profitable because arm's-length contracting sheds liability. Contracting is coercive because buyers are concentrated. Buyer concentration is sustained by intellectual monopoly. Monopoly profits are protected from taxation by transfer pricing. The whole system runs at speed because logistics has compressed circulation time, and it runs cheaply because ecological costs and reproductive labour are not paid for. Remove any one component and the others weaken. That is what makes it a system, and that is why single-issue reform, a code of conduct here, a minimum wage rise there, has so consistently failed to change the aggregate distribution. 6. Sectoral Cases 6.1 Fast fashion: the garment chain The garment industry is the classic buyer-driven chain and the clearest illustration of the mechanisms above. Structure. A handful of global retailers and brands sit at the top. Beneath them are first-tier suppliers, typically large factories in Bangladesh, Vietnam, Cambodia, Indonesia, India, Turkey, Ethiopia, and elsewhere. Beneath those are second and third tiers, smaller and often unregistered units, and beneath those, homeworkers. Entry barriers at the supplier level are low: sewing machines are cheap, skills are learned quickly, and capital requirements are modest. This is precisely why supplier margins are thin and buyer power is overwhelming. The squeeze. The #fast_fashion model, characterised by rapid design cycles, low prices, high volumes, and deliberate disposability, intensifies every pressure described in section 5. Lead times have collapsed. Order volumes are volatile. Prices paid per unit have been pushed down in real terms even as compliance costs have risen. Anner (2020, 2022) provides the evidence base for these claims, and his account of pandemic-era order cancellations remains the clearest demonstration of who holds power in the relationship. The human cost. The collapse of Rana Plaza in 2013, which killed more than eleven hundred workers, was the moment the model's logic became undeniable. The building had visible cracks. Workers were sent in anyway because an order had to ship. The subsequent Accord on Fire and Building Safety produced real improvements in structural safety in Bangladesh, and it should be acknowledged as such. But it did not touch wages, hours, or prices, which is to say it did not touch the mechanism generating the pressure. The ecological cost. Niinimaki et al. (2020) document the industry's environmental footprint in detail, including its enormous water consumption and the vast quantity of textile waste generated annually. Most of the pollution occurs at the production end, in dyeing and finishing, and most of the consumption occurs at the other end. This is #ecological_unequal_exchange in its most literal form. Brooks (2019) traces the further stage, the export of discarded Northern clothing back to Southern markets, where it undercuts local producers. Value distribution. The pattern in the #garment_industry is consistent across studies: the manufacturing country captures a small fraction of the retail price, with the great bulk absorbed by brand margin, retail margin, marketing, and distribution in the consuming country. The wage share of the retail price of a typical garment is customarily estimated in low single-digit percentages. The standard interpretation is that sewing is low-value work. The interpretation offered here is that sewing is where the value is produced, and the retail price records where it is captured. 6.2 Electronics: the smartphone chain The electronics chain differs in structure and confirms the argument in a different register. Structure. Here the barriers to entry at the assembly level are higher. Contract manufacturers are enormous firms in their own right, running facilities employing hundreds of thousands of workers. Yet the same asymmetry holds: the brand-owning firm captures the overwhelming share of the margin, while the contract manufacturer operates on margins of a few percent. The dormitory labour regime. Chan, Selden, and Pun (2020) provide the most thorough account of conditions inside the largest of these operations. The dormitory system is central to the analysis. Housing workers on site collapses the boundary between the working day and the rest of life, allowing production schedules to be extended at will, and permitting the recruitment of a young migrant workforce far from home, family, and any support network. Compulsory overtime, production quotas, military-style discipline, student "interns" drafted from vocational schools during peak production, and the wave of worker suicides that drew international attention are all documented in their study. The dormitory regime is the institutional form of #super_exploitation. It permits a wage below the cost of settled family reproduction, because the workforce is not settling and not reproducing itself locally. Reproduction happens in the countryside, unpaid, on someone else's account. Value distribution. Detailed teardown studies of earlier smartphone models found that the brand-owning firm captured well over half of the retail price as gross margin, while the labour cost of assembly amounted to a very small percentage of the retail price, in the low single digits. The components, from Japan, South Korea, Taiwan, Germany, and the United States, accounted for most of the rest. This distribution is normally presented as evidence that the assembly stage contributes little value. Read through the framework developed here, it is evidence of how much #surplus_value flows out of the assembly stage and is realised elsewhere as brand margin and #intellectual_monopoly rent. Beyond assembly. Crawford (2021) extends the analysis to the full material and labour infrastructure of digital technology, from mineral extraction through data labelling to logistics, arguing that the apparently immaterial digital economy rests on an intensely material and intensely exploitative substrate. Anwar and Graham (2021) and Woodcock and Graham (2020) show that the same logic now extends to digital services work, where platform intermediation has created a genuinely global labour market for tasks, with all the arbitrage possibilities that implies and none of the protections. 6.3 The extractive base: battery minerals Finally, a brief case at the base of the chain, where the language of #upgrading has the least credibility. The energy transition and the electronics industry both depend on minerals whose extraction is geographically concentrated. Cobalt is the emblematic case: the Democratic Republic of the Congo supplies the large majority of world output, and a significant share of that output comes from artisanal mining, work performed by hand with minimal equipment and no safety provision, often by people including children, in conditions documented by Kara (2023). The minerals enter the chain through traders and refiners, are processed largely outside the country of extraction, and emerge as components in products whose retail price bears no relation to the conditions of their origin. The DRC captures a tiny share of the final value of the products its minerals make possible. There is no plausible upgrading path from artisanal pit to battery design studio under current arrangements, and the resource is finite. Kara's account has been debated with respect to the precise proportion of output from artisanal sources and the prevalence of specific abuses, and readers should engage with that debate rather than accept any single figure uncritically. But the structural point is not seriously contested: the conditions of extraction and the distribution of the resulting value are grotesquely asymmetrical, and this asymmetry is the base on which the rest of the chain is built. 7. The Statistical Illusion: Why the Numbers Hide the Transfer If the argument above is correct, we should expect the process to be invisible in the official statistics. It is, and the reason is instructive. 7.1 Value added is not value created National accounts measure value added: the difference between the value of a firm's output and the value of its purchased inputs. Summed across an economy, this yields gross domestic product. The convention is universal and, for many purposes, useful. The problem is that it assumes the price at which a firm sells its output measures the value that firm created. If prices simply reflected labour times, this would be an acceptable approximation. But we established in section 3.3 that prices systematically diverge from values, redistributing #surplus_value across sectors and, where capital is mobile and labour is not, across borders. Once we accept that, "value added" becomes a measure of value captured, not value created. The consequence is that the accounts are constitutionally incapable of registering a transfer. If a supplier is squeezed to a two percent margin and a brand books a forty percent margin, the accounts record a low-productivity supplier and a high-productivity brand. The squeeze itself has vanished into the productivity statistics of the buying country. This is the #GDP_illusion, and it is the single most important reason why the transfer is so rarely discussed. 7.2 Trade in value added does not solve the problem The trade-in-value-added statistics developed over the last decade were designed to correct the double-counting produced by intermediate goods crossing borders repeatedly. They are a genuine improvement for that purpose. But they inherit the same underlying convention. They tell us more accurately where value added is recorded. They cannot tell us where value was produced, because they use the same prices to define both. 7.3 What would we see if we measured differently? The recent quantitative literature attempts precisely this. Hickel et al. (2021) ask what Southern exports would be worth if they were priced at Northern price levels, and interpret the difference as the drain. Ricci (2021) uses input-output tables to estimate the labour content embodied in trade and to compute net transfers on that basis. Carchedi and Roberts (2021) work from Marxian value categories directly. These approaches differ in their assumptions and give different magnitudes. What they share is a result: the direction of the net transfer is consistently from the #Global_South to the #Global_North, and the magnitude is large enough to be macroeconomically significant, on the order of trillions of dollars annually. When Hickel et al. (2022) note that the value of the resources net-appropriated from the South in a single year would be sufficient to end extreme poverty many times over, they are making a point about the scale of what is at stake. Reasonable people can and do dispute the counterfactual pricing assumptions. What is harder to dispute is that the standard accounts embed a counterfactual of their own, the assumption that market prices measure value creation, and that this assumption is never defended because it is never noticed. 8. Discussion: Rebuilding Dependency Theory 8.1 What must be discarded Some elements of classical dependency theory should be let go. The strongest version of the stagnationist thesis, that peripheral economies cannot industrialise at all under conditions of dependency, has been falsified. East Asian industrialisation happened. Chinese industrial capacity is now formidable. Manufacturing employment in the periphery has grown enormously. Equally, the nation-state is no longer an adequate unit of analysis. There is a capitalist class in Dhaka and a working class in Detroit. The Bangladeshi factory owner squeezing his workforce is not a victim of imperialism in the same sense as the worker on his line, even if he is himself squeezed by the buyer. Any framework that lumps them together as "the periphery" will produce bad politics as well as bad analysis. Kvangraven (2021) is right that the better dependency scholarship never made this mistake, but the popular version often did. 8.2 What must be retained What survives, and what the evidence in this paper supports, is the structural claim. The world economy is organised such that a systematic net flow of value runs from lower-wage to higher-wage zones, and this flow is not an accident of trade patterns but a consequence of the way capital, labour mobility, property rights, and market power are institutionally arranged. Peripheral industrialisation happened, but it happened in a form, subordinate, contract-based, dependent on foreign buyers, technologically shallow, that reproduces the transfer rather than ending it. The concept that best captures this is dependent industrialisation. The country builds factories. It does not build ownership of the assets that command the surplus. 8.3 What must be added Three things are new and require theoretical extension. First, the firm-level architecture. Classical dependency theory was written before the arm's-length supply chain existed in its current form. The lead firm that owns nothing and controls everything is a new object, and it must be at the centre of the theory. Suwandi (2019), Baglioni et al. (2020), and Palpacuer and Smith (2021) point the way. Second, #intellectual_monopoly. The centrality of intangible assets to contemporary #value_capture was not anticipated. Amin's imperialist rent gestured toward it; Durand, Milberg, and Rikap have specified it. Third, #social_reproduction. Feminist political economy has established that the value chain rests on a reproductive base that classical dependency theory largely ignored. Mezzadri's work makes it impossible to treat the wage relation as the whole story. 8.4 Confronting the objections An argument of this kind should meet its strongest critics directly. Four objections deserve serious treatment. Objection one: wage gaps reflect productivity gaps. The orthodox position is that Southern workers are paid less because they produce less per hour, given the capital and technology available to them. If true, this dissolves the entire argument. The response has two parts. First, physical productivity in export manufacturing is often not low at all. A modern garment factory in Vietnam or an electronics assembly plant in China runs at or near the global technological frontier for its process, with output per worker comparable to what the same process would yield anywhere. The wage gap in these facilities vastly exceeds any plausible physical productivity gap. Second, when productivity is measured in value terms, the argument becomes circular: the worker is deemed less productive because the output is assigned a lower price, and the output is assigned a lower price because the buyer has the power to set it. This is the circularity identified in section 7. The honest concession is that in some sectors and some countries, productivity differences are real and substantial, and not all of the wage gap is a transfer. The question is what share is, and that is an empirical question about which the estimates vary. Objection two: upgrading is real. Some suppliers do move up. Firms in Taiwan and South Korea moved from assembly to design to brand ownership. Chinese firms now compete at the technological frontier in several sectors. This is true and important. The response is that these cases share a common feature: powerful developmental states that protected domestic firms, directed credit, enforced technology transfer, and restricted foreign ownership, in a specific geopolitical window in which such policies were tolerated. The trade and investment regime constructed since the 1990s explicitly restricts exactly these policy instruments. The ladder that some countries climbed is being actively withdrawn, and Selwyn and Leyden (2022) show that the mainstream development literature recommending #global_value_chains integration does not address this. Objection three: China breaks the model. The rise of China, and the emergence of Chinese lead firms with their own supplier networks across Asia and Africa, complicates any simple #core_periphery map. Chinese capital now occupies a position in some chains that looks structurally similar to the position Northern capital occupies in others. This should be accepted rather than explained away. The theory is about structural positions in the world economy, not about a fixed list of countries. A country can move, and a country can contain both positions simultaneously. What the theory predicts is that the structure persists even as its occupants change, and that the value transfer continues to flow toward whoever controls the intangible assets and the finance. The rise of a new pole does not abolish the mechanism. Objection four: the quantitative estimates are unreliable. Critics of the drain literature argue, with some force, that the counterfactual pricing assumptions do a great deal of work, that the results are sensitive to specification, and that "what Southern exports would be worth at Northern prices" is not obviously the right question. This objection has real weight and should not be brushed aside. The appropriate response is epistemic humility about magnitudes combined with confidence about direction. Multiple independent methods, monetary input-output analysis, physical resource accounting, embodied labour accounting, and firm-level margin analysis, all point the same way. The convergence of methods that make different assumptions is the strongest evidence available. It does not license precision. 9. Political and Policy Implications If the analysis holds, what follows? 9.1 The limits of voluntary approaches Corporate social responsibility, private auditing, and voluntary codes of conduct have now been running for three decades. LeBaron (2020) reviews the evidence and finds the audit industry structurally incapable of detecting or correcting the abuses it was created to address, in part because it is paid for by the firms it audits, and in part because it targets the supplier while leaving the buyer's purchasing practices untouched. This is the key structural insight. An auditor can find excessive overtime in a factory. The auditor cannot find the order that made the overtime unavoidable, because the order was placed by the auditor's client. Any reform framework that regulates suppliers while exempting buyers is regulating the symptom. 9.2 Binding obligations on lead firms The logical alternative is to place binding legal obligations on the buying firm for conditions throughout its chain, including for the effects of its own purchasing practices. Mandatory due diligence legislation moves in this direction. Whether such legislation can be made effective, or whether it will become another compliance formality, is an open question and one of the more urgent research agendas in the field. 9.3 Wages and worker power The most direct route to reducing the transfer is to raise the price of Southern labour. This means unions, the right to strike, sectoral bargaining, and cross-border coordination among workers in the same chain. It also means confronting the fact that supplier states actively suppress wages as an export strategy, which is precisely the dynamic Marini identified. #labour_internationalism is not a slogan here but a structural requirement: because the chain is global, bargaining that is only national can be defeated by relocation. 9.4 Industrial policy and the question of delinking At the level of states, the implications point toward industrial policy, restrictions on profit repatriation, control over natural resources, regional integration among Southern economies, and reform or refusal of the intellectual property regime. Amin's language of #delinking is unfashionable but the underlying question, how does a country stop specialising in the part of the chain that gets squeezed, has not been answered by any other framework. 9.5 The fiscal and reparations question Finally, the analysis reframes debates about aid and about climate finance. If the North is a net recipient of value from the South, on the scale the recent estimates suggest, then transfers running the other way are not charity. They are, at best, partial restitution. Whether this argument is best framed in the language of #reparations or in the language of structural reform of the trading system is a political question. What the analysis establishes is that the framing of aid as generosity is empirically backwards. 10. Limitations Four limitations should be stated plainly. First, the paper synthesises secondary evidence and generates no new data. Its claims are only as strong as the sources it relies on, and several of those sources are contested. Second, the magnitude of #value_transfer cannot currently be measured with confidence. The paper argues for direction and order of magnitude, not for any specific figure, and readers should treat published point estimates with corresponding caution. Third, the analysis is weighted toward manufacturing chains. Services, agriculture, and finance operate differently and would require separate treatment. The extension of the argument to platform-mediated digital services work is sketched rather than developed. Fourth, the paper treats the #Global_South and #Global_North as analytical categories while acknowledging that they are internally divided by class, and it does not fully resolve the tension between a structural account organised around geography and a class account organised around ownership. That tension is real, and honest analysis should sit with it rather than paper over it. 11. Conclusion The modern supply chain is a remarkable technical achievement. It can move a design from a studio to a store in three weeks, across four continents, at a price the consumer barely notices. This paper has argued that it is also a machine for moving #surplus_value, and that the two facts are not separate. The mechanisms are identifiable and mundane. A wage set below what it costs to live. A contract that can be cancelled without penalty. A price pushed down while requirements are pushed up. A trademark that commands a margin it did not produce. A transfer price that relocates profit before it can be taxed. A shipping schedule that transmits every fluctuation in Northern demand directly onto the body of a Southern worker. A river that absorbs the cost of the dye. A woman who does thirty hours of unpaid work so that a sixty-hour wage can pretend to be sufficient. None of these is a scandal in the sense of being hidden. All of them are visible in the trade press, the corporate filings, and the audit reports. What is hidden is their combined result, and it is hidden by an accounting convention that records value where it is captured and calls that the place where it was made. Dependency theory, in the form it took fifty years ago, cannot simply be reinstated. Its unit of analysis was too coarse, and some of its predictions failed. But the question it asked, why does integration into the world market so reliably reproduce the subordination of the integrated, has not been answered by the frameworks that displaced it. The evidence assembled here suggests that the question was the right one, and that the answer lies in the architecture of the chain itself. The task for the next generation of research is to specify that architecture with more precision, to improve the measurement of transfer without pretending to a false exactness, and to identify the points in the system where organised pressure can actually change the distribution. The chain has choke points. So does the argument that justifies it. #global_value_chains #surplus_value #unequal_exchange #dependency_theory #super_exploitation #labour_arbitrage #imperialism #fast_fashion #electronics_manufacturing #monopsony #intellectual_monopoly #value_transfer #Global_South #political_economy #international_relations References Contemporary sources Anner, M. (2020). Squeezing workers' rights in global supply chains: Purchasing practices in the Bangladesh garment export sector in comparative perspective. Review of International Political Economy, 27(2), 320-347. doi:10.1080/09692290.2019.1625426 Anner, M. (2022). Power relations in global supply chains and the unequal distribution of costs during crises: Abandoning garment suppliers and workers during the COVID-19 pandemic. International Labour Review, 161(1), 59-82. doi:10.1111/ilr.12337 Anwar, M. A., and Graham, M. (2021). The Digital Continent: Placing Africa in Planetary Networks of Work. Oxford: Oxford University Press. Baglioni, E., Campling, L., Coe, N. M., and Smith, A. (Eds.). (2022). Labour Regimes and Global Production. Newcastle upon Tyne: Agenda Publishing. Baglioni, E., Campling, L., and Hanlon, G. (2020). Global value chains as entrepreneurial capture: Insights from management theory. Review of International Political Economy, 27(4), 903-925. doi:10.1080/09692290.2019.1657479 Brooks, A. (2019). Clothing Poverty: The Hidden World of Fast Fashion and Second-Hand Clothes (2nd ed.). London: Zed Books. Carchedi, G., and Roberts, M. (2021). The economics of modern imperialism. Historical Materialism, 29(4), 23-69. doi:10.1163/1569206X-12342992 Chan, J., Selden, M., and Pun, N. (2020). Dying for an iPhone: Apple, Foxconn, and the Lives of China's Workers. Chicago: Haymarket Books. Cobham, A., and Jansky, P. (2020). Estimating Illicit Financial Flows: A Critical Guide to the Data, Methodologies, and Findings. Oxford: Oxford University Press. Crawford, K. (2021). Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence. New Haven: Yale University Press. Dorninger, C., Hornborg, A., Abson, D. J., von Wehrden, H., Schaffartzik, A., Giljum, S., Engler, J.-O., Feller, R. L., Hubacek, K., and Wieland, H. (2021). Global patterns of ecologically unequal exchange: Implications for sustainability in the 21st century. Ecological Economics, 179, 106824. doi:10.1016/j.ecolecon.2020.106824 Durand, C., and Milberg, W. (2020). Intellectual monopoly in global value chains. Review of International Political Economy, 27(2), 404-429. doi:10.1080/09692290.2019.1660703 Fanning, A. L., and Hickel, J. (2023). Compensation for atmospheric appropriation. Nature Sustainability, 6, 1077-1086. doi:10.1038/s41893-023-01130-8 Foster, J. B., and Suwandi, I. (2020). COVID-19 and catastrophe capitalism: Commodity chains and ecological-epidemiological-economic crises. Monthly Review, 72(2), 1-20. Hickel, J., Dorninger, C., Wieland, H., and Suwandi, I. (2022). Imperialist appropriation in the world economy: Drain from the global South through unequal exchange, 1990-2015. Global Environmental Change, 73, 102467. doi:10.1016/j.gloenvcha.2022.102467 Hickel, J., Sullivan, D., and Zoomkawala, H. (2021). Plunder in the post-colonial era: Quantifying drain from the global South through unequal exchange, 1960-2018. New Political Economy, 26(6), 1030-1047. doi:10.1080/13563467.2021.1899153 Kara, S. (2023). Cobalt Red: How the Blood of the Congo Powers Our Lives. New York: St. Martin's Press. Khalili, L. (2020). Sinews of War and Trade: Shipping and Capitalism in the Arabian Peninsula. London: Verso. Kvangraven, I. H. (2021). Beyond the stereotype: Restating the relevance of the dependency research programme. Development and Change, 52(1), 76-112. doi:10.1111/dech.12593 LeBaron, G. (2020). Combatting Modern Slavery: Why Labour Governance is Failing and What We Can Do About It. Cambridge: Polity Press. Madariaga, A., and Palestini, S. (Eds.). (2021). Dependent Capitalisms in Contemporary Latin America and Europe. Cham: Palgrave Macmillan. Marini, R. M. (2022). The Dialectics of Dependency. New York: Monthly Review Press. Mezzadri, A. (2021). A value theory of inclusion: Informal labour, the homeworker, and the social reproduction of value. Antipode, 53(4), 1186-1205. doi:10.1111/anti.12701 Mezzadri, A., Newman, S., and Stevano, S. (2022). Feminist global political economies of work and social reproduction. Review of International Political Economy, 29(6), 1783-1803. doi:10.1080/09692290.2021.1957977 Ness, I. (2023). Migration as Economic Imperialism: How International Labour Mobility Undermines Economic Development in Poor Countries. Cambridge: Polity Press. Niinimaki, K., Peters, G., Dahlbo, H., Perry, P., Rissanen, T., and Gwilt, A. (2020). The environmental price of fast fashion. Nature Reviews Earth and Environment, 1, 189-200. doi:10.1038/s43017-020-0039-9 Palpacuer, F., and Smith, A. (Eds.). (2021). Rethinking Value Chains: Tackling the Challenges of Global Capitalism. Bristol: Bristol University Press. Patnaik, U., and Patnaik, P. (2021). Capital and Imperialism: Theory, History, and the Present. New York: Monthly Review Press. Ricci, A. (2021). Value and Unequal Exchange in International Trade: The Geography of Global Capitalist Exploitation. London: Routledge. Rikap, C. (2021). Capitalism, Power and Innovation: Intellectual Monopoly Capitalism Uncovered. London: Routledge. Rikap, C. (2022). Amazon: A story of accumulation through intellectual rentiership and predation. Competition and Change, 26(3-4), 436-466. doi:10.1177/1024529421992171 Rikap, C., and Lundvall, B.-A. (2021). The Digital Innovation Race: Conceptualizing the Emerging New World Order. Cham: Palgrave Macmillan. Selwyn, B., and Leyden, D. (2022). Oligopoly-driven development: The World Bank's Trading for Development in the Age of Global Value Chains in perspective. Competition and Change, 26(2), 174-196. doi:10.1177/1024529421995351 Sullivan, D., and Hickel, J. (2023). Capitalism and extreme poverty: A global analysis of real wages, human height, and mortality since the long 16th century. World Development, 161, 106026. doi:10.1016/j.worlddev.2022.106026 Torslov, T., Wier, L., and Zucman, G. (2023). The missing profits of nations. Review of Economic Studies, 90(3), 1499-1534. doi:10.1093/restud/rdac049 Woodcock, J., and Graham, M. (2020). The Gig Economy: A Critical Introduction. Cambridge: Polity Press. Foundational works cited for theoretical grounding Amin, S. (1976). Unequal Development: An Essay on the Social Formations of Peripheral Capitalism. New York: Monthly Review Press. Emmanuel, A. (1972). Unequal Exchange: A Study of the Imperialism of Trade. New York: Monthly Review Press. Marx, K. (1976 [1867]). Capital: A Critique of Political Economy, Volume One. London: Penguin Books. Marx, K. (1981 [1894]). Capital: A Critique of Political Economy, Volume Three. London: Penguin Books. Ricci, A. (2019). Unequal exchange in the age of globalization. Review of Radical Political Economics, 51(2), 225-245. doi:10.1177/0486613418773753 Selwyn, B. (2019). Poverty chains and global capitalism. Competition and Change, 23(1), 71-97. doi:10.1177/1024529418809067 Smith, J. (2016). Imperialism in the Twenty-First Century: Globalization, Super-Exploitation, and Capitalism's Final Crisis. New York: Monthly Review Press. Suwandi, I. (2019). Value Chains: The New Economic Imperialism. New York: Monthly Review Press.

This article studies Karl Marx's Capital, Volume One, not only as a work of economic theory but as a literary text, a translated text, and a teaching text. It brings together three fields that rarely speak to each other: literary and rhetorical analysis, translation studies, and global intellectual history. The first part argues that Marx wrote in a #gothic_metaphor register, borrowing vampires, werewolves, spectres, and animated objects from the popular horror writing of his century in order to make an abstract and invisible process feel physical to readers who had no training in #political_economy. The second part follows what happened to those images and to Marx's technical vocabulary when the book crossed language borders. It examines terms such as #Mehrwert (surplus value), Arbeitskraft (labour power), Wert and Wertform (value and value form), Verwertung (valorization), Verdinglichung (reification), and Entfremdung (alienation), and shows how choices made by translators quietly changed what regional readerships believed Marx had said. The third part traces circulation and teaching. It shows that movements in the #Global_South rarely read Capital as a whole. They read selected chapters, above all the chapter on the working day and the chapters on so-called primitive accumulation, because those chapters narrate violence, colonialism, and dispossession in a form that could be taught quickly in a reading circle or a party school. The article concludes that Capital's global authority rests less on complete comprehension than on the portability of a few metaphors and a few chapters, and it argues that students today should be taught to read Capital with the translation history visible rather than hidden. Keywords: Marx; Capital; rhetoric; gothic; translation studies; retranslation; surplus value; anti-colonial movements; reading circles; book history; pedagogy 1. Introduction Most students meet Capital as a difficult book about economics. They are told that it contains a theory of value, a theory of exploitation, and a theory of crisis, and they are usually asked to master the argument of the opening chapters before they are allowed to enjoy anything else. That approach hides something obvious. Capital is also a piece of nineteenth-century writing, produced in the same decades that produced sensational fiction, penny dreadfuls, factory inspectors' reports, gothic novels, and newspaper exposes of urban misery. Marx read all of these. He also wrote like a man who had read all of these. This article begins from a simple observation. When Marx wants to describe the thing that his readers cannot see, he does not reach for a diagram. He reaches for a monster. Capital is dead labour that lives by drinking living labour, in the manner of a vampire. The drive for extra hours of unpaid work is a hunger like that of a werewolf. The commodity, an ordinary wooden table, begins to dance and to produce strange ideas out of its wooden head. Value has a ghostly, phantom-like objectivity. Money and capital enter the world stained with blood and dirt. These are not accidental flourishes. They are a method of exposition, and they belong to a specific historical moment in European letters. The article's first claim, therefore, is rhetorical. Marx used the #gothic register because the object he was analysing had a genuinely uncanny quality: it was a social relation between people that appeared as a relation between things, and a system of exploitation that appeared as a free and equal exchange. Ordinary descriptive prose would have reported the appearance and lost the reality. Horror imagery let him show a world in which objects act and people are acted upon. This was a solution to a problem of #rhetoric, and it was also a solution to a problem of teaching. Marx wanted workers to read him, and workers were far more likely to have encountered a vampire story than a treatise by Ricardo. The second claim is about #translation_studies. A metaphor is not a decoration that can be lifted out and carried across a border unchanged. It is embedded in the sound, the register, and the connotations of a particular language. When Capital was translated into Russian, French, English, Japanese, Chinese, Spanish, Arabic, and dozens of other languages, translators had to decide what to do with words that were simultaneously technical and figurative. Some of these decisions flattened the horror into economics. Some of them produced new technical vocabularies that then became the standard scholarly language of an entire region. In East Asia, for example, the Marxist lexicon reached Chinese largely through Japanese compounds that had themselves been invented to catch German abstractions. Whole national traditions of Marxist scholarship were therefore built on a chain of decisions made by a small number of individual translators, most of whom are not household names. The third claim is historical and pedagogical. Books do not travel on their own. They travel through publishers, cheap paper, party schools, prison libraries, student unions, night classes, and reading circles. In the twentieth century, state publishing houses in Moscow and Beijing printed inexpensive editions in many languages and shipped them to Asia, Africa, the Caribbean, and Latin America. In many of these settings, few people read the whole of Volume One. What they read, and what they taught, was a selection. The chapters that travelled best were those that told a story: the chapter on the working day, with its factory inspectors and its child labour; the chapters on machinery; and above all the chapters on so-called #primitive_accumulation, which describe enclosure, colonial plunder, slavery, and the violent creation of a class with nothing to sell but its own labour power. For readers in colonised societies, that section did not read like history. It read like the news. Putting these three claims together gives us a way of describing Capital that is more accurate than either the economics-textbook version or the literary-appreciation version. Capital is a #textual_authority that functions differently in different places because it arrives in different words, in different fragments, and in different institutional settings. This article calls it a pedagogical artifact: an object that teaches, that is reshaped by teaching, and that carries its own history of reshaping inside it. The article is organised as follows. Section 2 reviews the relevant literature across the three fields. Section 3 sets out the theoretical framework. Section 4 describes the method and the materials. Section 5 analyses the gothic rhetoric of Capital in detail. Section 6 analyses translation, with particular attention to key terms and to the major English, Russian, French, East Asian, Spanish, and Arabic trajectories. Section 7 traces circulation and pedagogy in the Global South. Section 8 discusses what the three analyses together imply. Section 9 draws out practical consequences for students and teachers. Section 10 states the limitations. Section 11 concludes. 2. Literature Review 2.1 Marx as a writer The study of Marx's writing as writing has a long and uneven history. For most of the twentieth century, treating Capital as literature was viewed with suspicion by both sides. Orthodox readers feared that attention to style would soften the science. Literary readers often treated Marx as a source of theory rather than as an author with a style of his own. That situation has changed. The recent English publication of Ludovico Silva's study of Marx's literary style, originally written in Spanish, has reintroduced an argument that Marx's images are not ornaments but the load-bearing structure of his exposition (Silva, 2023). Silva insists that Marx thought through metaphor, and that trying to strip the metaphors away in order to reach a purely technical core misunderstands how the book works. Work on the specifically gothic and monstrous imagery has developed along two lines. Neocleous (2003) argued that Marx's vampire is a precise figure rather than a loose insult, because vampirism names a relation in which the dead feed on the living, which is exactly what Marx says capital does to labour. McNally (2011) extended this into a wider argument about the monstrous imagination of market societies, connecting Marx's monsters to bodily fear, to dissection and anatomy scandals, and to popular horror. Margaret Cohen's earlier notion of a Gothic Marxism, developed in relation to Walter Benjamin and surrealism, provides a name for a tradition in which #critical_reading of capitalism proceeds through the uncanny rather than around it (Cohen, 1993). Roberts (2017) takes a different literary route, reading Capital as a rewriting of Dante's Inferno, with a descent through circles of a social hell. Alongside this, a strong body of commentary aims to help readers through the difficulty of the text. Heinrich (2021) offers a chapter-by-chapter guide to the opening sections and is unusually careful about the German. Harvey (2018) provides a reading companion built out of decades of teaching. Jameson (2011) reads Volume One as a formal and narrative construction. Musto's edited volumes gather much of the current scholarship on the book's composition, its unfinished character, and its afterlives (Musto, 2019; Musto, 2020a). 2.2 Translation studies and political texts Translation studies has moved a long way from the idea that a translation is either faithful or unfaithful. Lefevere's account of translation as rewriting, shaped by patronage, ideology, and the poetics of the receiving culture, remains a foundation (Lefevere, 2017). Venuti's polemics against instrumental models of translation insist that translation always interprets, and that the fantasy of a neutral transfer of content is itself an ideological position (Venuti, 2019a, 2019b). His older distinction between domestication and #foreignization remains useful for describing what happens to Marx: some translators smooth him into the idiom of the receiving academic culture, others keep the German strangeness visible. The political dimension of translation has become a field in its own right. The Routledge handbooks on translation and politics and on translation and activism collect studies of how texts are moved by movements, and how movements are shaped by what they can read (Fernandez & Evans, 2018; Gould & Tahmasebian, 2020). Rundle's handbook of translation history provides methodological guidance for treating translation as a historical event with datable causes and consequences (Rundle, 2021). Baker's collection on unsettling translation gathers work that questions the neutrality of the translator's position (Baker, 2022). Casanova's model of a world literary space, structured by unequal centres and peripheries, helps explain why translations from German passed so often through French, English, and Russian before reaching other languages (Casanova, 2004). Specific studies of Capital's translation history are scattered rather than consolidated. Sutherland's essay on Marx's jargon is the best known intervention in English, and it turns on a single word, Gallerte, which he argues has been sanitised by English translators into something abstract when the German is disgustingly concrete (Sutherland, 2008). The publication of a new English translation of Volume One by Paul Reitter, edited by Paul North, has reopened the whole question in public, because it deliberately revisits many settled renderings and restores a rougher, stranger Marx (Marx, 2024). 2.3 Global circulation and anti-colonial reading The history of Marxism outside Europe has been rewritten in the last two decades. Anderson's work on Marx's late writings has shown that Marx himself became increasingly interested in non-European societies, in rural communes, and in colonial dispossession, which changes what it means to say that Capital was later applied to colonised societies (Anderson, 2025). Musto's biography of the last years shows the same restless late Marx (Musto, 2020b). Samaddar (2018) reads Marx from a postcolonial vantage point and argues that the categories of Capital were reworked, not simply imported, in colonial and postcolonial contexts. On Africa, the reissue of Walter Rodney's political essays under the title of a decolonial Marxism has made available his own account of how Marxist theory was taught and adapted at the University of Dar es Salaam (Rodney, 2022; Rodney, 2018). Cabral's writings on resistance and decolonisation show a leader constantly translating class analysis into the vocabulary of a peasant society (Cabral, 2016). Nkrumah's account of neo-colonialism uses the language of capital export and monopoly (Nkrumah, 1965). Getachew's history of anticolonial worldmaking and Gopal's history of insurgent metropolitan politics both show that these were not passive receptions but active reconstructions (Getachew, 2019; Gopal, 2019). Taiwo (2022) offers a sharp critique of some current uses of the word decolonisation, and is worth reading against the grain of the others. On Asia, Karl (2020) and Lin (2021) provide accessible accounts of the Chinese revolutionary use of Marxist categories, including the way the vocabulary was localised. Walker (2016) analyses the Japanese theoretical tradition, which produced some of the most technically demanding readings of Capital anywhere in the world. Saito's work draws directly on the Japanese tradition and on the ongoing scholarly edition of Marx's manuscripts (Saito, 2023, 2024). On Latin America, Bosteels (2012) discusses the ways Marxist and psychoanalytic vocabularies were absorbed and transformed. Mariategui's essays remain the classic statement that socialism in the Americas could not be a copy of a European model (Mariategui, 1971). On the Arab world, the translated writings of Mahdi Amel show a theorist working carefully with Capital's categories while insisting that colonial societies have their own structure (Safieddine, 2021). 2.4 The gap Each of these literatures is strong. What is missing is a study that follows a single chain: from the metaphor in the German, to the decision made by a translator, to the classroom or reading circle where the translated chapter was taught, to the political conclusion drawn from it. This article attempts that chain. 3. Theoretical Framework Three ideas organise the analysis. First, metaphor as cognition, not decoration. Following Silva (2023) and the broader tradition of #metaphor_theory, this article treats Marx's images as instruments of thought. A metaphor selects a structure from one domain and imposes it on another. When Marx describes capital as dead labour that feeds on living labour, he imports from the vampire story a set of relations: the parasite requires the host; the parasite is animated only by what it takes; the host is drained gradually rather than killed at once; and the whole process happens at night, unseen. Each of these maps onto a claim in the theory. The metaphor is thus a compressed argument. Any translation that keeps the word vampire but loses the surrounding vocabulary of drinking, sucking, draining, and nightly return loses part of the argument. Second, translation as rewriting under constraint. Following Lefevere (2017) and Venuti (2019a), this article assumes there is no neutral transfer. Every translator works under constraints: the vocabulary available in the target language, the expectations of the receiving academic or political culture, the patron who pays, the censor who permits, and the political organisation that wants a usable text. Because Capital is both a scientific work and a political weapon, its translators were pulled in two directions. A translator who wanted the book to be respectable in a university might reduce the horror. A translator who wanted the book to move a movement might increase it. Both are rewritings, and both leave traces. Third, circulation and selection as the real unit of reception. Following the methods of book history and #circulation_of_texts, this article assumes that what matters is not what a book says but what parts of it were printed, priced, distributed, assigned, summarised, and repeated. A chapter that is photocopied and taught a thousand times has more historical effect than a chapter that is admired and skipped. This is why the article gives sustained attention to which chapters travelled. Taken together, these three ideas produce the article's central hypothesis: the global authority of Capital is the product of a portable rhetoric, a plural set of translations, and a selective #curriculum, and the three reinforce each other. The most portable parts of the book are the parts with the strongest images and the clearest narrative, and those are precisely the parts that were selected for teaching, and therefore the parts that most needed translating well and were most often translated badly. 4. Materials and Method This is a qualitative, comparative, and historical study. It does not use statistical methods. It uses three procedures. Close reading of the source text. The analysis of imagery works from the German of Volume One, in the editions consolidated by the ongoing historical-critical edition of Marx's and Engels's works, and from the major published German versions of 1867, 1872 to 1873, and 1883. Attention is given to words whose figurative charge is easy to miss: ungeheuer at the very opening, gespenstisch in the section on value, Gallerte in the analysis of abstract labour, Werwolfsheisshunger in the chapter on the working day, and Blutsauger and Vampyr where capital confronts the worker. This is a #close_reading exercise, not a quantitative corpus study, although a corpus study would be a valuable next step. Comparative analysis of translations. The article compares renderings of a small set of key terms and images across the authorised French version prepared by Joseph Roy and revised heavily by Marx himself, the English version by Samuel Moore and Edward Aveling supervised by Engels, the later standard English version by Ben Fowkes, the recent English version by Paul Reitter edited by Paul North, and, through secondary scholarship, the Russian, Japanese, Chinese, Spanish, and Arabic traditions. This is a #comparative_analysis at the level of the term and the image rather than a full collation of every sentence. Historical tracing of circulation and teaching. The final section works from published histories, memoirs, and the writings of movement intellectuals to reconstruct which parts of Capital were taught, in what settings, and to what end. The evidence here is uneven by design, because the archives of #reading_circles and night schools are thin. Where the record is thin, the article says so. A note on the writer's position is appropriate. This study works directly with German, English, French, and Spanish materials. For the Russian, Japanese, Chinese, and Arabic traditions it depends on secondary scholarship and on translated primary sources. That dependence is a limitation, discussed in Section 10, and readers with those languages will be able to correct and extend the argument. 5. Capital as Nineteenth-Century Literature: The Gothic Register 5.1 The problem Marx had to solve Before looking at the monsters, it is worth stating the problem they solve. Marx's claim is that in a society of commodity production, the fundamental relations between people take the form of relations between things. The worker and the employer meet as free and equal parties in the market. Nobody is whipped. Nobody is legally owned. The contract is voluntary. And yet, Marx claims, a surplus is extracted, systematically, from the worker's day. The extraction happens not in the visible sphere of exchange, where everything looks fair, but in the hidden sphere of production, and it is measured in time rather than in goods. This creates a communicative problem of the first order. The thing that has to be described is by definition not visible on the surface, and the surface actively contradicts it. Everything a reader can see suggests fairness. If Marx simply asserted that unfairness lies beneath, he would be one more moralist. If he wrote only in the language of equations and definitions, he would reach a few hundred specialists. The gothic solves this. Gothic writing is precisely the genre of the invisible force that acts on bodies, of the dead that will not stay dead, of the inanimate that moves, of the respectable house with a horror in the cellar. Marx's readers, whether educated or not, knew this grammar. By writing capital as an animated corpse, an insatiable appetite, a spectral objectivity, and a monstrous automaton, Marx could make the systemic and abstract feel bodily and immediate. This is a pedagogical decision as much as a stylistic one. 5.2 The vampire The vampire is Marx's most famous figure and his most exact. In the analysis of the working day, capital is described as dead labour which comes to life only by sucking living labour, and which lives the more, the more of that living labour it drinks. Every element of this is theoretically loaded. Dead labour, in Marx's system, means the past labour that is congealed in machines, buildings, and raw materials. It is a technical term. On its own it is inert. It contributes value to the product only to the extent that it is used up, and it cannot create new value by itself. Living labour, the activity of workers in the present, is the only source of new value. So the machine, which appears to be productive, is in fact a corpse; and the corpse is animated only when a living body is applied to it. That is exactly the structure of the vampire story: the undead thing walks, but only on borrowed blood. The image also carries a temporal claim. The vampire does not consume the victim in a single meal. It returns, night after night, taking a little at a time, and the victim weakens without being able to name the cause. Marx's argument about the working day has the same shape. The extension of the day by an hour, the shortening of the meal break by ten minutes, the encroachment on the night, the appropriation of the hours of sleep: this is a slow, cumulative predation, and this is why Marx piles up the reports of the factory inspectors, which read like a long list of small bleedings. #dead_labour and #living_labour are thus not merely categories, they are characters in a horror narrative. Finally, the image carries a moral claim without stating a moral. Marx does not say the capitalist is a wicked man. He repeatedly says the opposite: the individual capitalist is the personification of a function, driven by competition. The vampire figure allows him to describe a compulsion, a hunger, rather than a vice. Vampires are not evil in the way that a cruel employer is evil. They are hungry in a way that cannot be reasoned with. That distinction is the whole difference between a moral critique of capitalists and a structural critique of #capital_accumulation, and the metaphor carries it. 5.3 The werewolf Less discussed but equally precise is the werewolf. In the chapter on the working day, Marx describes capital's ravenous appetite for surplus labour with a compound that fuses the werewolf with a violent hunger. The German compound, Werwolfsheisshunger, is one word, and it hits hard. Where the vampire figures parasitism, the werewolf figures loss of control and periodic transformation. The werewolf is a person by day and a beast by night. It cannot help what it becomes. It has no memory of restraint. This maps onto Marx's argument about limits. Capital, considered as a self-expanding value, has no internal principle that says enough. It cannot decide to be moderate. Left alone, it will extend the working day until the worker is destroyed, because the compulsion to accumulate is not a psychological trait but a condition of survival in competition. Only an external force, in Marx's account the collective action of workers and the resulting factory legislation, can impose a limit. Notice what this does pedagogically. It converts an argument about structural compulsion, which is hard, into an image of an appetite that cannot stop, which is easy. And it prepares the political conclusion: you do not reason with a #werewolf_hunger, you chain it. 5.4 The dancing table and the fetish The most philosophically dense passage in Volume One is the section on the fetish character of the commodity, and it is also the passage where Marx's literary technique is most visible. He takes an ordinary wooden table. As a useful object, he says, there is nothing mysterious about it. But the moment it appears as a commodity, it changes into something strange: it stands on its head and, out of its wooden brain, brings forth grotesque ideas. Two contexts make this passage sharper than it looks. The first is religion: fetishism was, in the anthropological writing of Marx's time, the name for the worship of objects that people themselves had made. Marx is saying that commodity society reproduces this structure at the heart of the most modern and rational economy. The second context is the craze for spirit-rapping and table-turning that swept Europe and North America in the 1850s, in which participants sat around tables that supposedly moved and communicated. Marx's dancing table is a joke about that craze, and the joke has a point. Educated people who laughed at the spiritualists nevertheless believed that prices had lives of their own, that markets were confident or nervous, that money worked. Marx's move is to say: your table dances too, and you do not even notice. The related phrase gespenstische Gegenstaendlichkeit, usually rendered in English as phantom-like or spectral objectivity, is the point where the horror and the theory become inseparable. Value, in Marx's account, is real, in that it governs behaviour, prices, and lives, but it is not a physical property of any object. You cannot find it with a microscope. It is a social relation that appears in the form of a thing. That is, precisely, a ghost: something that has effects without having a body. #phantom_objectivity is therefore not a poetic flourish added to the concept of value; it is the concept of value. This has a direct implication for translation. Any translator who renders the phrase as an abstract or ordinary expression, such as an unsubstantial or merely formal objectivity, saves the reader from confusion and destroys the argument at the same time. 5.5 Gallerte, or the horror of abstract labour In the opening chapter, Marx has to explain what it means to say that commodities as values are nothing but congealed human labour, considered without regard to the specific form of that labour. The German uses the word Gallerte. In technical German this is a jelly, and in the food industry of the period it named the gelatinous mass produced by boiling down bones, tendons, and animal remains. It is not a clean, scientific word. It is a word from the slaughterhouse and the glue factory. Sutherland (2008) made this the centre of an argument about English translation. Rendered as congealed quantities of homogeneous human labour, the sentence sounds like physics. Rendered with the connotations of the German word intact, it sounds like a description of human beings boiled down into a residue in which all individual difference has been destroyed. That is exactly what abstract labour is: the reduction of the specific, skilled, embodied activity of a particular person to an indifferent quantity of human effort in general, measured by time. The gothic point is that this reduction is not a mental operation performed by an economist. It is performed by the market, on real people, every day. #Gallerte is therefore one of the clearest cases where a single lexical choice determines whether a reader experiences the theory of value as a piece of accounting or as a piece of horror. The recent English retranslation restores the visceral sense with a rendering closer to a gelatinous mass, and the difference in the reader's experience is considerable (Marx, 2024). 5.6 The automaton, the factory, and the monster of machinery In the chapters on machinery and large-scale industry, the imagery shifts. The vampire drinks; the werewolf devours; but the factory system is a single vast mechanism into which human beings are inserted as parts. Marx describes the automatic factory as a mechanical monster whose limbs fill whole buildings, and whose demonic power is hidden behind the regular and apparently calm movement of its parts. He inverts the ordinary relation between worker and tool: in handicraft, the worker uses the tool; in the factory, the machine uses the worker. Critics have often connected this to Mary Shelley's creature, and the connection is instructive even though Marx does not build his argument on that novel. What the two share is the structure of a creation that escapes its creator and turns on him. Marx's version is more radical, because his monster is not a single artificial body but an entire social system, and because in his account nobody in particular built it and nobody in particular can stop it. This section of the book also contains one of Marx's most brutal images, in which the movement of capitalist progress is compared to a crushing wheeled idol that grinds its worshippers beneath it. The reference would have been familiar to British readers from colonial travel writing about India, and this is significant for the argument of Section 7: even inside the metropolitan text, the vocabulary of horror is already partly a colonial vocabulary. 5.7 Blood and dirt: the gothic history of accumulation The last part of Volume One turns from analysis to history. Marx asks how the whole arrangement came into being: how did there come to be, on one side, owners of money and means of production, and on the other side, people with nothing to sell but themselves. The official story, which he mocks as a nursery tale, is that some people were diligent and thrifty and others were lazy and spent everything. He calls it so-called original accumulation, and the so-called is doing heavy work, because his point is that the phrase is a lie. His counter-history is written in blood. It includes the enclosure of common land, the clearing of estates, the criminalisation of the dispossessed through bloody legislation against vagrancy, the plunder of the Americas, the enslavement and trade of African people, the conquest of India, and the conversion of the colonies into hunting grounds. It ends with the famous statement that capital comes into the world dripping from head to foot, from every pore, with blood and dirt. Formally, this is the gothic origin story. The respectable house has a corpse under the floorboards. The wealth of the present is haunted by a founding crime that has been officially forgotten and that keeps returning. #colonial_violence is not, in this account, an unfortunate episode in the history of capitalism. It is the condition of its birth. The rhetorical structure here is worth naming because it explains the chapter's later career. These pages are narrative, concrete, dated, and geographical. They mention places that readers in Asia, Africa, the Caribbean, and Latin America could recognise as their own. They require no prior grasp of the value form. They can be read on their own. Section 7 will show that this is exactly what happened. 5.8 Why the gothic worked To summarise the argument of this section: Marx's monsters are not literary decoration laid over an economic skeleton. They are a solution to the problem of representing a system that is real but invisible, coercive but contractual, and destructive but impersonal. They perform four functions at once. They make the abstract sensory. They locate agency in the system rather than in individuals, since a vampire's hunger is a condition, not a choice. They organise time, because predation is repeated and cumulative. And they carry a political conclusion, since monsters are not persuaded but restrained or destroyed. This also explains why the imagery is unevenly distributed. It is thickest where the argument is either most abstract, as in the fetish section, or most brutal, as in the working day and the history of accumulation, and thinnest in the technical passages on the forms of value and on the rate of surplus value. Marx used horror where he needed to convert either abstraction or atrocity into something a reader could feel. 6. Translation: What Happens When the Monsters Cross Borders 6.1 There is no single original Any account of translating Capital must begin with an uncomfortable fact: the German text is not stable. Marx published the first edition in 1867. He then rewrote substantial parts, including the crucial opening on the value form, for the second German edition. He then supervised a French translation by Joseph Roy, published in instalments in the early 1870s, and he revised that version so heavily that he said in his own note to readers that it possessed a scientific value independent of the original and should be consulted even by readers who knew German. After his death, Engels edited further German editions and supervised the first complete English translation. The consequences are large and are still being worked out by the scholarly edition of the complete works, which prints the different versions side by side. Some passages exist in a simpler French form and a denser German form. Later translators into other languages had to choose which base to use. Some translated from the German. Some translated from the French, because French was the more accessible international language in their region. Some translated from the Russian or from the English, and thus from a translation of a translation. This means that the phrase the original German text, used casually in seminars, is misleading. When a reader in Buenos Aires, Hanoi, or Cairo disagreed with a reader in Berlin about what Marx said, the disagreement was sometimes not about interpretation at all. They were reading different books. Recognising this is the first step toward an honest #source_text and #target_text analysis of the tradition. 6.2 The terms that carry the weight Six clusters of terms concentrate most of the translation problems. Mehrwert. The compound joins mehr, meaning more, and Wert, meaning value, and it names the value produced by the worker beyond the value of the worker's own labour power. English settled early on surplus value, and this rendering spread widely, including into the calques used in East Asia. It is a workable choice, but it has a cost. Surplus, in English economic usage, suggests something left over, an excess, a happy remainder. The German compound is blunter and closer to more-value, a value added on. The English term subtly makes the extracted value sound like a residue rather than like the whole purpose of the operation. In languages where the term was rendered as excess or leftover value, the same softening occurred, and students had to be taught to un-learn the everyday sense of the word before they could use it technically. #surplus_value is therefore a good example of a translation that is not wrong and is nevertheless slightly off-key. Arbeitskraft. Marx's central innovation in the analysis of the wage is the distinction between labour and labour power. The worker does not sell labour, which is an activity; the worker sells the capacity to work for a period of time, and it is the gap between what that capacity costs and what it can produce that generates surplus value. The German Kraft carries the sense of force, strength, and physical power, as in the vocabulary of nineteenth-century physics. English labour power and French force de travail preserve this. Some early translations into other languages blurred the distinction, using a single word for both labour and the capacity to labour. Where that happened, the core of the theory of exploitation became almost impossible to state, and commentators had to invent the distinction again in their own prefaces. #labour_power and #Arbeitskraft are thus not a scholastic nicety; they are the hinge of the argument. Wert and Wertform. The theory of the #value_form is the hardest part of the book and the part that Marx rewrote most anxiously. He was trying to show how the value of a commodity can only be expressed in the body of another commodity, and how money emerges from this necessity. Translating this requires a target language that can hold a difference between value, exchange value, and the form of value, and that can bear the strange syntax in which one commodity expresses its value in another. Translators frequently simplified. The French version that Marx himself approved simplified. Every simplification makes the passage more readable and slightly less able to do what Marx wanted it to do, which is to show that money is not a convenient invention but a necessary form of a social relation. Verwertung. This word names the process by which value is used to make more value, the self-expansion of capital. English lacked a natural equivalent, and for a long time it was translated with vague phrases about the creation of value or the increase of value. The later adoption of the term valorization gave English readers a technical word, at the cost of introducing a term that means nothing outside Marxist economics. The point for our purposes is that for several generations, English readers of the Moore and Aveling version did not have a single stable word for one of the book's central processes, which made the process harder to see. #Verwertung is a case where the absence of a term in the receiving language delayed the reception of an idea. Verdinglichung, Versachlichung, and Entfremdung. Here a common student error must be corrected, and the correction matters for this article's argument. Alienation, Entfremdung, is not the central term of Capital. It is the central term of the manuscripts Marx wrote in 1844, which were not published until 1932 and which reached wide readerships in French and English only in the late 1950s and early 1960s. Their appearance changed global Marxism, because they gave readers a humanist and philosophical Marx to set against the official economic one. In Capital itself, the corresponding work is done by the analysis of the fetish, and by the vocabulary of thingification and objectification, that is, the process by which relations between people take on the form of relations between things. The translation history here is unusually consequential. Because the 1844 manuscripts arrived late, and arrived in translation, whole national debates about whether there was a young humanist Marx and an old scientific Marx were conducted on the basis of texts whose availability was an accident of archival history. The question of #Entfremdung in the Global South was in many places a question about a book that had been published a century after it was written and had crossed two or three languages on the way (Musto, 2021). The gothic vocabulary. Finally there are the words analysed in Section 5. Vampyr, Blutsauger, Werwolfsheisshunger, gespenstisch, Gallerte, ungeheuer. These are the terms where translators face the sharpest choice. Keep the horror and risk sounding unscientific, or normalise it and lose the argument. The historical tendency, in most languages and for most of the twentieth century, was to normalise. Marxism was fighting for scientific respectability, often inside universities and often inside states, and a text full of monsters was an embarrassment. It is only in the last two or three decades that translators and editors have begun to restore the strangeness deliberately. 6.3 The English trajectory English is the best documented case and can stand as a model. The first complete English Volume One, by Samuel Moore and Edward Aveling under Engels's supervision, appeared in 1887. It is a Victorian text, and it reads like one. Its virtues are considerable: it is clear, it is idiomatic, and it was authorised by the man who knew Marx best. Its costs are also considerable. It smooths the German. It divides Marx's German chapters into a larger number of English chapters, so that the German eighth chapter on the working day becomes the English chapter ten, and the German twenty-fourth chapter on so-called original accumulation becomes an entire English part running across several chapters. This is not a small matter for the argument of Section 7, because when a movement in Africa or Asia assigned Chapter 26 to 33 for a study group, it was assigning a unit that Marx had not created; the unit was made by his English translators. The Fowkes version, published in the 1970s, is the standard scholarly English text for most readers today. It is more rigorous, more consistent in terminology, and it introduced or fixed several key terms in English, including valorization. It restores much of the German architecture. It is also, by design, more academic in tone, and some of the coarser physical force of the German remains muted. The Reitter translation, edited by North and published in 2024, is the first major English retranslation in nearly fifty years, and it makes a point of returning to the strangeness (Marx, 2024). It restores concrete and unpleasant vocabulary where earlier translators reached for abstraction, and it is explicit about the fact that the smoothing of Marx's language was an interpretive act with political consequences. Whatever one thinks of its individual choices, its existence proves the article's thesis: a #retranslation is an argument about what the book is. 6.4 Russian, French, and the mediating languages The first foreign translation of Capital was into Russian, appearing in 1872, the work of several hands, most prominently German Lopatin and Nikolai Danielson. The censor allowed it, according to a much repeated account, because it was judged too difficult to be dangerous. The judgment was wrong, and it is a useful lesson in the sociology of #mistranslation and misreading: the state read the book's difficulty as a guarantee of harmlessness, when in fact the book's difficulty guaranteed that it would be summarised, excerpted, and taught, and that its summaries would travel far faster than its argument. The French version by Roy, revised by Marx, matters for a different reason. Because French was the international language of educated readers across much of Africa, the Middle East, Southeast Asia, and Latin America, a great many people encountered Capital first in French. Frantz Fanon and Ho Chi Minh, among many others, worked in French. This means that Marx's own simplifications, made for a French readership he considered less philosophically trained, were passed on to readers far beyond France. When we ask what Marx said, we should sometimes ask which Marx reached which continent. Casanova's model of unequal literary space applies exactly here (Casanova, 2004). German was the language of the original, but German was not the language of world circulation. Ideas moved through Paris, London, and later Moscow. The result was a chain of translations in which small distortions accumulated, and in which peripheral readerships were dependent on the interpretive decisions of intermediary metropolitan translators whom they never met. 6.5 East Asia, and the making of a lexicon The East Asian case is the clearest demonstration that translation can create a discipline. In Japan, an intense engagement with Marxian economics developed from the 1910s onward. A complete Japanese translation of Volume One was achieved in the early 1920s, and Marxian economics soon acquired a strong institutional presence in Japanese universities, producing a theoretical tradition, later associated with the school around Uno Kozo and its successors, which subjected the value-form chapters to a level of scrutiny rarely matched elsewhere (Walker, 2016). Japanese scholarship has also been central to the modern historical-critical edition of Marx's manuscripts, and the recent revival of interest in Marx's ecological notebooks owes much to it (Saito, 2023). To translate Marx, Japanese scholars had to invent words. They built compounds out of Chinese characters to carry German philosophical and economic abstractions. These compounds then travelled to China, where a large part of the modern Chinese vocabulary of the social sciences, including the terms for surplus value, capital, class, and production relations, derives from these Japanese coinages. This is one of the great and under-taught facts of global intellectual history. The Chinese Marxist lexicon is, at its root, a translation of a translation. #Japanese_neologisms shaped what a Chinese reader could think about Marx before that reader opened the book. In China, partial translations circulated in the 1920s and 1930s, and the first complete Chinese translation of the three volumes, by Guo Dali and Wang Yanan, was published in 1938, in the middle of war. From then on, Capital in Chinese was not primarily a university text. It was a text of political formation, taught in party schools and study sessions, and read in the light of a revolution being fought largely by peasants rather than by an industrial proletariat. That context shaped which chapters mattered and how they were read (Karl, 2020; Lin, 2021). 6.6 Spanish, Portuguese, and Latin America Spanish-language readers had access to a translation of Volume One from the end of the nineteenth century, and two later versions became decisive. The version prepared by Wenceslao Roces and published in Mexico in the 1940s carried Capital into a Spanish-speaking world in which many of the leading readers were exiles from the Spanish Civil War, and it became for decades the standard. The later version prepared by Pedro Scaron for a major Latin American publisher in the 1970s was made with close attention to the differences between the German editions and the French version, and with an explicit concern for a Latin American readership rather than a Spanish one. The disputes between these versions were not merely philological. They were arguments about who owned the text, about whether Latin American Marxism should receive its concepts through Madrid or make them for itself, and about how technical the language of a revolutionary text should be. #Spanish_translation history in the region is thus a chapter in the history of intellectual independence. 6.7 The Arab world and mediated reception In the Arab world, the reception of Capital was for a long time mediated. Many of the most important theorists read Marx in French or English, or in Russian, and the earliest Arabic renderings of major Marxist texts appeared well after the book's political influence in the region had begun. Soviet publishing houses produced Arabic editions and distributed them widely from the middle of the twentieth century, which meant that for many readers the Arabic Marx arrived pre-interpreted, packaged with introductions written in Moscow. What is striking is how much original theoretical work was done nonetheless. The writings of Mahdi Amel, now available in English, show a thinker working carefully with the categories of Capital while refusing to treat colonial societies as simply backward versions of European ones, and insisting that the colonial relation produces a specific mode of production rather than a delayed copy of the metropolitan one (Safieddine, 2021). Samir Amin's work on accumulation on a world scale performs a comparable operation. #Arabic_reception is therefore a case where mediation did not prevent originality; the theorists worked around the gaps in what had been translated. 6.8 What translation does to metaphor Bringing the strands together, three patterns recur. Flattening. The most common fate of Marx's images is to be reduced to their conceptual content. The horror is treated as a rhetorical wrapper and quietly removed, especially in translations produced for or by state institutions and universities, where the priority was to present Marxism as a science. The consequence is a Marx who is harder to read and less memorable, and whose central arguments about the inversion of persons and things become abstract propositions rather than experiences. Terminological hardening. In the opposite direction, translation frequently converts a word that was flexible in German into a fixed technical term in the target language. Once a term is fixed, it becomes a password. Students learn to use it correctly rather than to think with it. This is how a living metaphor becomes a piece of #jargon, and it is one reason why so many readers find Marxist writing repetitive. Re-gothicisation. Recently, and in several languages at once, translators and editors have begun deliberately to restore the strangeness of the German, on the argument that the horror was the theory. The new English Volume One is the most visible instance (Marx, 2024). This should be understood as the current phase of an old struggle, not as a final correction. The next generation will make its own choices, and it will do so under its own political constraints. The lesson for students is not that earlier translators were incompetent. Moore, Aveling, Roy, Fowkes, Roces, Scaron, Guo, Wang, and their counterparts elsewhere were serious people working under pressure and often at personal risk. The lesson is that a translation is a reading, and that the reading which a reader inherits arrives already made. To read Capital critically is therefore to read the translation critically, which almost no undergraduate curriculum currently teaches anyone to do. 7. Circulation and Pedagogy in the Global South 7.1 The material conditions of reading Ideas travel on paper, and paper has a price. Any honest account of how Capital reached the Global South must begin with publishing and distribution rather than with philosophy. From the 1930s and especially after 1945, state publishing houses in the Soviet Union and later in China produced enormous quantities of Marxist classics in many languages, printed on cheap paper and sold at prices far below commercial cost, often subsidised to the point of being nearly free. These editions travelled to India, to West and East Africa, to the Caribbean, to Latin America, and to Southeast Asia, and they were sold in party bookshops, at student fairs, in trade union offices, and by mail. For many readers, the physical Capital they held was a Moscow edition with a preface explaining what they were about to read. #Progress_Publishers and their equivalents were among the most effective educational institutions of the twentieth century, and they were not universities. This matters for our argument in two ways. First, the editions came with an interpretive frame attached. A preface is a set of instructions for reading. Second, the cheapness of these editions meant that Capital was accessible to people who could not have afforded a scholarly edition, which is exactly the population Marx said he wanted to reach. The cost of that accessibility was interpretive dependence. Alongside the state publishers, national left publishing houses grew: in India, publishers linked to the communist movement issued translations into Bengali, Hindi, Malayalam, Marathi, and other languages across the twentieth century; in West Africa and the Caribbean, small presses and party journals published extracts and commentaries; in Latin America, university presses and independent left publishers produced editions and study guides. Extracts, pamphlets, and summaries outnumbered complete editions by a wide margin, and this is the central fact of the reception. 7.2 Selection: which chapters were taught, and why Here we reach the heart of the historical argument. Capital, Volume One, is long, and its opening chapters are the hardest. The value form, the money form, the fetish, the general formula of capital: this is demanding material, and it defeats most first-time readers, including most first-time readers in Europe. Political movements that needed to educate cadres in a matter of weeks, often in the intervals of clandestine work, could not begin there. What they used instead were the narrative chapters. Three clusters dominate. The working day. This chapter, English chapter ten, is a documentary. It quotes the reports of factory inspectors, describes the deaths of young workers, catalogues the theft of minutes at meal breaks, and follows the political struggle over the legal limitation of hours. It requires no prior theory. It is enraging. And it ends with a clear political lesson: the limit on exploitation was not granted, it was won by organised force. For a trade union education class anywhere in the world, this chapter is close to ideal. #the_working_day became one of the most reprinted single chapters of Marx anywhere. Machinery and large-scale industry. This cluster mattered wherever industrialisation was arriving, or was being planned by a new state. It offered a way of thinking about technology that neither worshipped it nor rejected it, and it gave newly independent states a language for asking who controls the machines. #machinery_and_industry was widely taught in the developmentalist decades. So-called primitive accumulation. This is the section that made Capital an anti-colonial book. It is history, it is bloody, it names colonies, slavery, and conquest, and it can be read on its own. For readers in societies that had been dispossessed, this section was not an account of the European past. It was an account of their present. When Marx writes that the discovery of gold and silver in the Americas, the enslavement and burial alive of the indigenous population in mines, and the conversion of Africa into a commercial hunting ground for enslaved people signalled the rosy dawn of capitalist production, he provides a sentence that could be, and was, printed on a leaflet. The consequence of this selection is important and rarely stated plainly. The Capital that shaped the twentieth-century Global South was, in the majority of cases, a Capital without the value form. It was a book about #enclosure, exploitation, machinery, and empire, and not a book about the logic of the commodity. This does not mean the readers were wrong. It means the book they read was, in effect, a different book from the one taught in most European philosophy departments, and their disagreements with European Marxists were often disagreements between two different anthologies. 7.3 Africa: Dar es Salaam, Rodney, Cabral, Nkrumah The University of Dar es Salaam in the 1960s and 1970s was one of the most important sites of Marxist teaching anywhere in the world. Students and staff from across the continent and the diaspora argued about class, race, imperialism, and the peasantry, in a newly independent country that was attempting a socialist path. Walter Rodney taught there, and his essays from that period, now collected, show what teaching Marx in that setting actually involved: constant translation of European categories into African conditions, constant argument about whether those categories fitted, and a refusal to abandon them simply because they did not fit neatly (Rodney, 2022). Rodney's best known book, on how Europe underdeveloped Africa, is in an important sense an extension of Capital's chapters on original accumulation into the twentieth century and onto the African continent. It takes the argument that capitalism was born from plunder, and asks what happened to the places that were plundered. That is a pedagogical relationship as much as a theoretical one: it is what happens when a brilliant teacher takes the chapter that his students find most immediately convincing and follows it as far as it will go. #Walter_Rodney is thus a case study in productive selective reading. Amilcar Cabral offers a different case. Leading a liberation war in a largely peasant society, he faced a problem: the European vocabulary of proletariat and bourgeoisie described almost nobody in Guinea. His response was not to abandon class analysis but to rebuild it, analysing the actual strata of colonial society and the position of the small educated group from which the leadership came, and insisting that this group had to commit what he called class suicide by identifying with the peasantry. That argument is unthinkable without Marx and unthinkable as a simple application of Marx. #Amilcar_Cabral shows theory being remade in the field. Kwame Nkrumah's writings on neo-colonialism take the categories of capital export and monopoly and turn them into an account of how formal independence could coexist with continued economic control (Nkrumah, 1965). His account of the philosophical basis of African socialism attempted something even harder, namely to connect Marxist materialism with African communal traditions. #Kwame_Nkrumah is a reminder that the reception of Capital in Africa was always simultaneously a debate about African intellectual traditions. Frantz Fanon deserves mention here even though he was not primarily a reader of Capital. His insistence that Marxist analysis must be stretched when it is applied to the colonial situation, because in the colony the economic base and the visible fact of race are not separable in the European manner, is the single most influential sentence in the history of #anti_colonial_movements reading Marx. He read in French, in a French Marxist milieu, and his stretching is also, in part, a response to what French Marxism had made of Marx (Shatz, 2024). #Frantz_Fanon is thus a translation event as much as a theoretical one. 7.4 Latin America: Mariategui, Cuba, and the fight over categories Jose Carlos Mariategui, writing in Peru in the 1920s, put the problem with unusual sharpness. He argued that socialism in Latin America could not be a tracing or a copy of a European model but had to be a heroic creation, built from the actual conditions of the continent, and he took seriously the communal traditions of indigenous Andean society as a possible basis for a socialist future (Mariategui, 1971). This is exactly the question the late Marx himself was asking about the Russian commune, though Mariategui could not have known that, because those letters and drafts were not widely available at the time (Anderson, 2025). This is a striking instance of the translation problem at its deepest level. A question Marx had explored in his final years, in notebooks and letters that would not be published or translated for decades, had to be independently reinvented in Peru because the relevant texts had not travelled. #Jose_Carlos_Mariategui was, without knowing it, redoing work already done, because the archive was slow. In Cuba after 1959, Capital became a state pedagogical instrument, taught in schools, party institutions, and workplaces. Che Guevara's critical notes on political economy, written in this period and published long afterwards, show a leader arguing with the standard Soviet manuals and returning to Marx's own text in order to dispute the way the categories were being applied to a transitional economy (Guevara, 2008). This is a case where reading Capital directly, rather than through the official summaries, was itself a political act inside a socialist state. #Che_Guevara illustrates that the tension between the text and its authorised interpretation did not disappear after a revolution succeeded; it intensified. 7.5 Asia: study circles, party schools, and the peasant question In China, the study of Marxist classics was organised systematically in party schools from the revolutionary period onward. Study sessions worked through selected texts, with cadres reading and discussing under guidance. The circumstances made certain choices inevitable. The Chinese revolution's social base was peasant, and Capital analyses an industrial society; the most immediately usable material was therefore the historical and the political rather than the analysis of the commodity. The categories of Capital were also filtered through a Soviet interpretive apparatus and through the Japanese-derived lexicon discussed above (Karl, 2020; Lin, 2021). #Yanan_study_circles are a model of what movement pedagogy looks like: short texts, guided reading, collective discussion, and immediate application. In India, the situation was different again. English was widely available among the educated, so the Moore and Aveling text and later the Moscow English editions circulated directly, while translations into Indian languages built regional Marxist vocabularies. In Kerala, a dense culture of libraries, reading rooms, and study circles made left political education unusually widespread. Indian Marxist debate then produced one of the most sophisticated regional bodies of work on precisely the questions that Capital left open for colonial societies: the nature of agrarian relations, the character of a colonial mode of production, and whether the categories of Volume One applied to a countryside that had been reshaped by empire (Samaddar, 2018). In Vietnam, as in much of French-colonised Asia and Africa, the entry was through French, in Paris as much as in Hanoi. Ho Chi Minh's political formation happened in the French metropolitan left, and the texts he read were French texts. Again the shape of the empire determined the shape of the reading list. 7.6 The Arab world The Arab reception, as noted, was heavily mediated, but this did not make it derivative. The urgent theoretical question in the region was the relation between class and nation, and between anti-imperialist struggle and social revolution. Mahdi Amel's insistence that the colonial relation constitutes a distinct structure, rather than a temporary delay on the path to a European destination, is a direct response to a way of using Capital that treats Europe as the model and everywhere else as behind (Safieddine, 2021). #Mahdi_Amel and #Samir_Amin both, in different registers, refused the developmentalist reading, and both did so from within the categories of Marx's critique rather than by abandoning them. 7.7 A general pattern Across all these regions, a pattern repeats. First, the text arrives in translation, often in a translation made elsewhere, and often with an introduction that instructs the reader. Second, only part of the text is taught, and the part that is taught is chosen for its narrative power and its immediate political usefulness rather than for its place in the logical order of the argument. Third, the fit between the categories and the local society is poor, and this misfit becomes productive. The most creative Marxist theory in the twentieth century was produced by people who found that Marx's categories did not fit their societies and who refused either to abandon the categories or to force the society into them. Fourth, the imagery travels better than the analysis. The vampire, the bloodstained birth of capital, the monstrous machine: these images crossed languages, cultures, and levels of education with far less loss than the value form. A slogan can carry a metaphor. It cannot carry a dialectical derivation. This last point closes the circle with Section 5. Marx wrote in gothic images because he needed to make an invisible system visceral for readers without technical training. That rhetorical choice is precisely what allowed his book to become a global political text a century later, in languages he never read, among readers he never imagined. The rhetoric was built for portability, and it proved portable. 8. Discussion: Capital as a Pedagogical Artifact The three analyses converge on a single description. Capital is not best understood as a fixed body of doctrine that was more or less accurately received in various places. It is better understood as an artifact that was remade at each stage of its journey, and whose remaking is legible if we look at the right evidence. At the level of rhetoric, Marx built a text with two speeds. There is a slow, technical, and demanding argumentative spine, concentrated in the analysis of value, money, and capital. And there is a fast, sensory, narrative surface, concentrated in the historical and descriptive chapters and in the recurring images of monstrosity. The two are not separable in principle, since the images carry the arguments. But they are separable in practice, and the history of the book's reception is largely the history of that separation. At the level of translation, each language received the fast surface more easily than the slow spine. Metaphors of blood, hunger, and haunting survive translation reasonably well, because they draw on bodily experience that is widely shared. Terminological systems do not survive translation well, because they depend on fine distinctions that a target language may not have and may not want. The result, repeated across dozens of languages, is a text whose horror arrived intact and whose logic arrived damaged. At the level of pedagogy, teachers everywhere responded rationally to this situation. If the narrative chapters are gripping, translatable, and immediately relevant, and the analytical chapters are difficult, poorly rendered, and abstract, then a teacher with limited time and a class of workers, students, or cadres will teach the narrative chapters. Over decades, that rational local choice produced a global outcome: a worldwide Marxism organised around dispossession, exploitation, and imperialism, rather than around the commodity form. This is not a story of decline or of betrayal. It is a story of how a book becomes an institution. But it does have consequences, and two are worth naming. The first is that many long-running disputes inside Marxism are partly artefacts of translation and selection. Debates about whether Marx was a humanist or a scientist, about whether alienation is central or marginal, about whether Capital applies to agrarian and colonised societies: all of these were conducted by people reading different portions of a shifting corpus in different languages at different times. Recognising this does not dissolve the disputes, but it changes their character. Some of them are not disagreements about the world. They are disagreements about which book we are talking about. The second is that the current revival of interest in Marx's own manuscripts, driven by the historical-critical edition and by scholars working across German, Japanese, and English, is not merely philological housekeeping. It has already changed the picture substantially, particularly regarding Marx's late interest in ecology, in non-European societies, and in communal forms of property (Saito, 2023; Anderson, 2025; Foster, 2020). A reception history that was shaped by what happened to be published and translated is being partly rewritten now that more of the archive is available. Students entering the field today are, for the first time, in a position to read a Marx that is closer to the manuscripts than any previous generation, and they should be told that this is a historical novelty rather than the normal state of affairs. 9. Implications for Students and Teaching This article is written for students, and it should end with something usable. Six practical points follow from the analysis. Read the translator's preface. It is not a formality. It tells you which German edition was used, what the translator decided about key terms, and often what the translator thought the book was for. If a preface does not tell you these things, that silence is itself information. Learn a handful of German words. Not the language, just the words: Mehrwert, Arbeitskraft, Wert, Wertform, Verwertung, Verdinglichung, Gallerte, gespenstisch. Knowing that a single German word lies under two different English words, or that two German words lie under a single English word, is enough to keep you alert. This is the cheapest possible investment in #critical_reading. Read at least one passage in two translations. The fetish section is ideal, because it is short, famous, and translated many times. Reading the dancing table passage in an older and a newer version, side by side, teaches more about translation than any amount of theory. Do not skip the images. When Marx writes vampire, ask what the vampire is doing in the argument. The image is usually a compressed statement of a relation you are about to be shown at length. Students trained to treat metaphor as decoration read past the thesis and then complain that they cannot find it. Know which chapters you have actually read. Most people who say they have read Capital have read the first chapter, the chapter on the working day, and the section on primitive accumulation. That is a legitimate and historically respectable way to read the book, but it is a selection, and you should know that you have made it. Treat the history of the book as part of the book. The fact that the first translation was Russian, that the French edition was revised by Marx himself, that the English chapter numbering is a translator's invention, that the Chinese vocabulary came through Japanese, and that whole regions read the book through Moscow prefaces, is not background trivia. It is the record of how a text becomes powerful. #book_history is not an optional supplement to theory; for a text like this one, it is theory. For teachers, one further point. If the historical evidence shows that movements taught the narrative chapters first and the analytical chapters later or never, that is worth taking seriously rather than lamenting. There is a strong case for teaching Capital backwards: begin with the history of dispossession, move to the working day, and only then return to the commodity, at which point students will have a reason to care about the value form because they will already know what it is for. #critical_pedagogy of Marx has been quietly doing this for a century, without saying so. 10. Limitations Four limitations should be stated plainly. First, linguistic reach. The author works directly in German, English, French, and Spanish. The claims about Russian, Japanese, Chinese, and Arabic translations rest on secondary scholarship and on translated primary sources. Specialists in those languages will be able to correct, qualify, and enrich this account, and some of the generalisations offered here will not survive that contact. Second, the evidentiary base for pedagogy is thin by nature. Reading circles, night classes, prison study groups, and party schools rarely leave archives. Much of what we know comes from memoirs, which are shaped by hindsight, and from the published writings of unusually articulate participants, who are by definition not typical. The claim that certain chapters dominated teaching is well supported by the pattern of reprinting, excerpting, and citation, but it is not supported by systematic classroom records, because those records mostly do not exist. Third, this article treats Volume One almost exclusively. Volumes Two and Three, edited by Engels from manuscripts, have their own translation and reception histories, and the questions raised here apply to them with additional complications, since the editing of those volumes was itself an interpretive intervention. Fourth, the article's central metaphor claim, that gothic imagery aided global portability, is an argument from plausibility and from the pattern of what was quoted, excerpted, and reprinted. It would be strengthened considerably by a quantitative study of how often specific images appear in pamphlets, posters, speeches, and textbooks across languages. That is a project this article recommends rather than performs. 11. Conclusion Capital is a book about invisible things, and Marx knew that the hardest part of his task was not proving his case but making his readers see it. His solution was to write the invisible as a haunting. Capital is dead labour that drinks the living. Its appetite for unpaid hours is the hunger of a beast that cannot stop itself. Value is a ghost with real effects. The commodity, that most ordinary object, gets up and dances. And the whole system was born, not in thrift and diligence, but in enclosure, conquest, and slavery, dripping from head to foot with blood and dirt. Those images did what he needed them to do, and they did more. They survived the crossing into languages Marx never studied, in editions he never saw, among readers he never anticipated. A dialectical derivation of the money form does not survive a bad translation. A vampire does. When Capital reached the colonised world, the chapters that spread were the chapters that told stories, and above all the chapters on original accumulation, which described to readers in Asia, Africa, the Caribbean, and Latin America the history of their own dispossession and told them that this history was not an accident but a foundation. The consequence is a paradox that should be taught rather than hidden. The global authority of Capital rests only partly on the argument that Marx considered his greatest achievement, and rests heavily on the parts he wrote to make that argument bearable. The rhetoric that was supposed to be the servant of the theory turned out to be the vehicle that carried the theory across the world, and sometimes the vehicle arrived without its passenger. For students, the lesson is not cynical. It is practical. Read the translation as well as the text. Read the images as arguments. Know which chapters you have read and which you have skipped, and ask who made that selection for you and why. Capital is a #pedagogy as much as it is a theory, and it has been teaching, and being reshaped by teaching, for more than a century and a half. To read it well is to read it as what it has become: a #nineteenth_century_literature masterpiece, a chain of translations, and a curriculum, all at once. References Anderson, K. B. (2025). The late Marx's revolutionary roads: Colonialism, gender, and indigenous communism. Verso. Baker, M. (Ed.). (2022). Unsettling translation: Studies in honour of Theo Hermans. Routledge. Bosteels, B. (2012). Marx and Freud in Latin America: Politics, psychoanalysis, and religion in times of terror. Verso. Cabral, A. (2016). Resistance and decolonization (D. Wood, Trans.). Rowman and Littlefield. Callinicos, A., Kouvelakis, S., & Pradella, L. (Eds.). (2021). Routledge handbook of Marxism and post-Marxism. Routledge. Casanova, P. (2004). The world republic of letters (M. B. DeBevoise, Trans.). Harvard University Press. Cohen, M. (1993). Profane illumination: Walter Benjamin and the Paris of surrealist revolution. University of California Press. Fanon, F. (2004). The wretched of the earth (R. Philcox, Trans.). Grove Press. Fernandez, F., & Evans, J. (Eds.). (2018). The Routledge handbook of translation and politics. Routledge. Foster, J. B. (2020). The return of nature: Socialism and ecology. Monthly Review Press. Foster, J. B., & Clark, B. (2020). The robbery of nature: Capitalism and the ecological rift. Monthly Review Press. Getachew, A. (2019). Worldmaking after empire: The rise and fall of self-determination. Princeton University Press. Gopal, P. (2019). Insurgent empire: Anticolonial resistance and British dissent. Verso. Gould, R. R., & Tahmasebian, K. (Eds.). (2020). The Routledge handbook of translation and activism. Routledge. Guevara, E. (2008). Critical notes on political economy: A revolutionary humanist approach to Marxist economics. Ocean Press. Harvey, D. (2018). A companion to Marx's Capital: The complete edition. Verso. Heinrich, M. (2019). Karl Marx and the birth of modern society: The life of Marx and the development of his work, Volume 1. Monthly Review Press. Heinrich, M. (2021). How to read Marx's Capital: Commentary and explanations on the beginning chapters (A. Locascio, Trans.). Monthly Review Press. Jameson, F. (2011). Representing Capital: A reading of Volume One. Verso. Karl, R. E. (2020). China's revolutions in the modern world: A brief interpretive history. Verso. Lefevere, A. (2017). Translation, rewriting, and the manipulation of literary fame (2nd ed.). Routledge. Lin, C. (2021). Revolution and counterrevolution in China: The paradoxes of Chinese struggle. Verso. Mariategui, J. C. (1971). Seven interpretive essays on Peruvian reality (M. Urquidi, Trans.). University of Texas Press. Marx, K. (2024). Capital: Critique of political economy, Volume 1 (P. Reitter, Trans.; P. North, Ed.). Princeton University Press. McNally, D. (2011). Monsters of the market: Zombies, vampires and global capitalism. Brill. Musto, M. (Ed.). (2019). Marx's Capital after 150 years: Critique and alternative to capitalism. Routledge. Musto, M. (Ed.). (2020a). The Marx revival: Key concepts and new interpretations. Cambridge University Press. Musto, M. (2020b). The last years of Karl Marx: An intellectual biography. Stanford University Press. Musto, M. (2021). Karl Marx's writings on alienation. Palgrave Macmillan. Neocleous, M. (2003). The political economy of the dead: Marx's vampires. History of Political Thought, 24(4), 668-684. Nkrumah, K. (1965). Neo-colonialism: The last stage of imperialism. Thomas Nelson and Sons. Prashad, V. (2020). Washington bullets: A history of the CIA, coups, and assassinations. Monthly Review Press. Roberts, W. C. (2017). Marx's Inferno: The political theory of Capital. Princeton University Press. Rodney, W. (2018). The Russian Revolution: A view from the Third World (R. D. G. Kelley & J. J. Benjamin, Eds.). Verso. Rodney, W. (2022). Decolonial Marxism: Essays from the Pan-African revolution. Verso. Rundle, C. (Ed.). (2021). The Routledge handbook of translation history. Routledge. Safieddine, H. (Ed.). (2021). Arab Marxism and national liberation: Selected writings of Mahdi Amel (A. Giordani, Trans.). Haymarket Books. Saito, K. (2023). Marx in the Anthropocene: Towards the idea of degrowth communism. Cambridge University Press. Saito, K. (2024). Slow down: The degrowth manifesto. Astra House. Samaddar, R. (2018). Karl Marx and the postcolonial age. Palgrave Macmillan. Shatz, A. (2024). The rebel's clinic: The revolutionary lives of Frantz Fanon. Farrar, Straus and Giroux. Silva, L. (2023). Marx's literary style (P. Brito Nunez, Trans.). Verso. Sutherland, K. (2008). Marx in jargon. World Picture, 1. Taiwo, O. (2022). Against decolonisation: Taking African agency seriously. Hurst. Venuti, L. (2019a). Contra instrumentalism: A translation polemic. University of Nebraska Press. Venuti, L. (2019b). Theses on translation: An organon for the current moment. Flugschriften. Walker, G. (2016). The sublime perversion of capital: Marxist theory and the politics of history in modern Japan. Duke University Press. Hashtags #Das_Kapital #Karl_Marx #GothicMarxism #TranslationAndPower #GlobalMarxism #MarxStudies #CapitalVolumeOne #PostcolonialTheory #ComparativeLiterature #HistoryOfIdeas #TextualCirculation #PoliticalEconomyEducation #CriticalPedagogyStudies #HumanitiesResearch #STULIB

This paper asks a simple but difficult question: what happens to Marx's value theory when thinking machines enter the workshop? Using the categories of Capital, the study treats #artificial_intelligence systems, advanced robotics and large computing infrastructures as forms of #constant_capital, that is, as transferred value rather than as new value, and it treats human workers as the only source of #variable_capital and therefore the only source of #surplus_value. From this starting point the paper builds a conceptual and formal argument. If firms replace workers with machines across the economy, the #organic_composition_of_capital rises. If the rate of exploitation does not rise fast enough to compensate, the general #rate_of_profit must fall. The paper develops this claim in four steps. First, it clarifies why AI, despite its apparent "intelligence", cannot be classified as #living_labour in Marx's sense. Second, it presents a formal model with worked numerical illustrations showing the conditions under which AI adoption accelerates or delays the #falling_rate_of_profit. Third, it examines the powerful #counteracting_tendencies that operate in the AI era: cheapening of machinery, the enormous mass of hidden and low-paid human labour behind AI systems, the extension of working time through always-on digital tools, the global relocation of production, and above all the capture of #rent by a small number of firms that control models, data and compute. Fourth, it reads recent empirical literature on the labour share, profit rates, productivity growth and the energy cost of computation to see whether the theoretical expectations find support. The paper concludes that AI does not abolish the #labour_theory_of_value; instead it stretches, hides and globally redistributes human labour while raising #capital_intensity. The likely medium-term outcome is not the disappearance of value production but a highly unequal economy in which a few #intellectual_monopoly firms enjoy extraordinary returns, the majority of capitals face compressed profitability, and the pressure to intensify #exploitation grows. The paper closes with implications for students, researchers and policy, and with a research agenda for testing these propositions empirically. Keywords: labour theory of value; organic composition of capital; artificial intelligence; automation; rate of profit; constant capital; digital capitalism; political economy of technology 1. Introduction 1.1 The problem Every generation of new machines produces the same two reactions. Some people say the machines will free humanity from work. Others say the machines will destroy livelihoods. Both reactions were present when the power loom arrived, when the assembly line arrived, when the computer arrived, and both are present again with the arrival of large-scale #machine_learning systems. What is missing from most of this debate is a serious question about value. Not "will machines take jobs?" but "where does profit come from if machines do the work?" Karl Marx gave an answer that remains uncomfortable for both defenders and critics of capitalism. In his account, machines do not create new value. They transfer the value already contained in them, piece by piece, into the products they help make. Only human #labour_power, bought as a commodity and set to work for longer than it takes to reproduce its own value, produces a surplus. From this simple premise a paradox follows. Each individual capitalist has a strong incentive to replace workers with machines, because the firm that produces more cheaply than its competitors captures extra profit. But if every capitalist does the same thing, the system as a whole removes the very source of the surplus it lives on. The individual rational act becomes a collective irrational outcome. This is the core of what Marx called the law of the #tendency_of_the_rate_of_profit_to_fall. Artificial intelligence pushes this paradox to its sharpest point yet. Earlier machines replaced the muscles of workers. AI claims to replace their judgement, their language, their pattern recognition, and increasingly their creative output. If a machine can write the report, draft the code, answer the customer, and inspect the weld, then the last refuge of irreplaceable human labour appears to be shrinking. Some writers conclude that Marx's theory is now obsolete, because value is produced by algorithms and data rather than by people. Others conclude the opposite, that Marx's theory is now more relevant than ever, because it alone explains why a technologically dazzling economy can also be a stagnating and unequal one. 1.2 Research aim and questions This paper takes the second position, but it does so critically and with attention to the many objections raised against Marx's law. The aim is not to defend Marx as a matter of loyalty. The aim is to test whether his categories still cut reality at the joints when applied to an economy organised around models, data centres and #automation. Four research questions organise the study. Conceptually, should AI systems be classified as #constant_capital or as something new that Marx's categories cannot hold? Formally, under what conditions does the diffusion of AI raise the #organic_composition_of_capital fast enough to depress the general rate of profit, and under what conditions do #counteracting_tendencies dominate? Empirically, do recent trends in the labour share, capital intensity, profitability and productivity support the theoretical expectation? Politically and socially, what follows for workers, students and policymakers if the analysis holds? 1.3 Contribution The contribution of this paper is threefold. First, it offers a careful classification argument. Much of the existing debate assumes rather than demonstrates that AI belongs to constant capital. This paper shows why the classification holds even when a machine appears to "think", and it identifies the precise point at which the classification would break down, namely the emergence of an artificial agent that is a legal seller of its own labour power in a labour market. That point has not been reached and may never be reached, but naming it clearly makes the theory falsifiable rather than merely rhetorical. Second, the paper provides a transparent formal treatment with numerical illustrations that a student can reproduce with a calculator. Much writing in this field is either purely philosophical or purely econometric. The middle ground, where a clear model shows exactly which parameter must move in which direction, is thinly occupied. Third, the paper insists on a fact that both AI enthusiasts and some Marxists overlook. AI is not labour-free. It rests on an enormous and geographically dispersed mass of #living_labour: data annotators, content moderators, warehouse workers, delivery drivers, electricians, miners of cobalt and lithium, construction crews for data centres, and the users whose activity generates the training material. The apparent disappearance of labour in the rich countries is partly a relocation and a re-description of labour, not its abolition. Any serious value analysis must follow the labour wherever it goes. 1.4 Structure of the paper Section 2 reviews the relevant literature. Section 3 sets out Marx's conceptual framework in accessible language. Section 4 explains the method. Section 5 addresses the classification of AI as constant capital. Section 6 develops the formal model of the organic composition and profitability. Section 7 analyses counteracting tendencies. Section 8 examines the hidden labour behind AI. Section 9 discusses rent and intellectual monopoly. Section 10 reviews empirical signals. Section 11 discusses scenarios and synthesises the argument. Section 12 draws implications. Section 13 states limitations. Section 14 concludes. 2. Literature Review 2.1 Three research traditions The literature relevant to this paper flows in three broad streams that rarely meet. The first is mainstream economics of #automation. Here the central concepts are tasks, skills and displacement. Acemoglu and Restrepo (2020) show that the introduction of industrial robots in United States commuting zones reduced employment and wages in exposed local labour markets, with each additional robot per thousand workers associated with measurable job losses. Their later work (Acemoglu and Restrepo, 2022) attributes a large share of the rise in wage inequality since 1980 to the automation of tasks previously performed by workers without college degrees. Acemoglu and Johnson (2023) broaden the argument historically, insisting that technology does not automatically raise shared prosperity and that the direction of technological change is a political choice. Acemoglu (2025) offers a deliberately sober macroeconomic estimate of generative AI's likely effect on total factor productivity over a decade, arguing that the aggregate gains will be modest rather than transformative. Brynjolfsson (2022) warns of a "Turing trap", the tendency of firms and researchers to pursue human imitation rather than human augmentation, a choice that concentrates income in the hands of capital owners. This tradition is rich in evidence but weak in value theory. It measures the movement of employment and wages without asking where profit comes from. It treats capital as a productive factor that generates income by its own nature, which is precisely the assumption Marx contested. The second stream is critical political economy of digital technology. Crawford (2021) traces the material substrate of AI through minerals, energy, labour and data, refusing the myth of the cloud as immaterial. Pasquinelli (2023) reconstructs a long history of machine intelligence as the codification and appropriation of collective human knowledge and the division of labour, arguing that algorithms are, in a precise sense, condensed social relations. Steinhoff (2021) provides one of the most direct engagements with our problem, examining the AI industry through the categories of labour and capital and warning against the assumption that automation liberates. Dyer-Witheford, Kjosen and Steinhoff (2019) had already asked whether AI represents an "inhuman power" capable of dispensing with the human worker, concluding that capital's drive to eliminate #living_labour is real but self-undermining. Muldoon, Graham and Cant (2024) document the human labour that feeds the machine, from annotation to moderation. Altenried (2022) describes the digital factory in which platform work and algorithmic management extend, rather than replace, the industrial organisation of labour. Delfanti (2021) shows in detail how, inside Amazon warehouses, robots do not remove workers but reorganise and intensify their work. The third stream is Marxist value theory and crisis theory proper. Carchedi and Roberts (2023) restate and defend the law of the tendency of the rate of profit to fall as the organising principle of twenty-first century capitalism, applying it to finance, technology and climate. Moseley (2023) offers a close reading of value theory in the opening of Capital. Heinrich (2021) provides a widely used pedagogical reconstruction of Marx while remaining sceptical about whether the tendency of the rate of profit to fall has the status of a law. Basu, Huato, Jauregui and Wasner (2023) reconstruct world profit rates over six decades and find a long declining trend with partial recoveries, providing the kind of evidence the theory needs. Tsoulfidis and Tsaliki (2019) reconstruct classical political economy in a modern framework and defend the relevance of a falling profitability tendency for understanding long waves. 2.2 The gap The gap is visible as soon as the three streams are laid side by side. Mainstream economics measures automation but does not theorise value. Critical technology studies theorise power and labour but rarely produce a formal value model. Marxist crisis theory produces the model but usually treats AI as one more machine, without examining the specific features of learning systems: their non-rival information content, their extreme fixed cost and near-zero marginal cost of reproduction, their fast #moral_depreciation, and their dependence on a global tail of low-paid annotation labour. This paper aims to occupy exactly that gap. It links the value model to the specific technical and institutional features of AI and to the empirical record. 2.3 The "Fragment on Machines" controversy No literature review on this topic can avoid the passage in the Grundrisse usually called the Fragment on Machines, in which Marx imagines a stage where wealth depends less on direct labour time than on the general state of science and technology, the #general_intellect. Autonomist readings, from the Italian workerist tradition onward, treat this passage as a prophecy of a knowledge economy in which the law of value collapses. Post-work and post-capitalist writers have used it to argue that measurement of value by labour time has already broken down. There are two serious problems with this reading. First, the Fragment is a notebook entry written more than a decade before the published Capital, and Marx's mature theory does not repeat its conclusion. In Capital he insists that as long as commodities are exchanged and labour power is bought, value continues to be measured in #socially_necessary_labour_time, however much science is embodied in the machinery. Second, the empirical claim in the autonomist reading is weaker than it appears. If value were no longer measured by labour time, firms would not be so obsessed with labour costs, working hours, productivity metrics and outsourcing. They are obsessed with these things precisely because the measurement still operates. Pasquinelli (2023) and Steinhoff (2021) both offer readings of the general intellect that keep it inside, rather than outside, the value form. This paper follows them. 3. Conceptual Framework: Marx's Categories in Plain Language 3.1 Value, use value and socially necessary labour time For Marx, a commodity has two faces. It is useful, and it is exchangeable. Its usefulness, its use value, is qualitative and cannot be compared across different goods: a chair is not more or less useful than a loaf of bread in any measurable sense. Yet commodities are exchanged in definite proportions, which means they share a common substance that can be measured. That substance is abstract human labour, and its measure is #socially_necessary_labour_time, the time required to produce a good under the average conditions of productivity and average intensity prevailing in society. Two points about this definition are constantly misunderstood by students and by critics. First, it is socially necessary, not individually necessary. A slow worker does not create more value by working slowly. If a new technology halves the average time required to make a shirt, the value of every shirt falls, including shirts already made by old methods. Technical progress therefore destroys value even as it multiplies physical output. Second, value is not price. Prices oscillate around values and, in the aggregate and after the formation of a general rate of profit, around prices of production. The theory does not predict that any given item sells for exactly its labour content. It claims that the total surplus available for distribution as profit, interest and #rent is produced by surplus labour, and that the distribution of that total among firms and sectors follows the movement of capital seeking the highest return. 3.2 Constant capital and variable capital Marx divides the capital advanced in production into two parts, and the division is the hinge on which the entire argument turns. #constant_capital, written as c, is the value laid out on means of production: raw materials, energy, buildings, tools, machines, and, in our case, servers, chips, data centres, software licences and trained models. It is called constant because it does not change its magnitude in the production process. It transfers its existing value to the product, no more and no less, either all at once (raw materials) or gradually through depreciation (#fixed_capital). #variable_capital, written as v, is the value laid out on the purchase of #labour_power, the human capacity to work. It is called variable because it does change magnitude. The worker is paid the value of their labour power, which is roughly the cost of the goods and services needed to reproduce them and their family at the customary standard. But the worker can work longer than the time needed to produce that value. The extra labour, performed but not paid, is #surplus_value, written as s. The formula for the value of a commodity is therefore: W = c + v + s where W is total value, c is transferred value, v is the reproduced value of labour power, and s is the new value produced above that. Notice what this implies. Machines contribute to use value, and they can contribute enormously. A robot can produce a thousand units where a worker produced ten. But machines contribute to exchange value only by handing over what was already put into them by past labour. This is why Marx calls machinery #dead_labour: it is labour already performed, congealed in a physical form, which now confronts the living worker as capital. 3.3 The rate of surplus value and the organic composition Two ratios follow. The rate of surplus value (or rate of exploitation) is: e = s / v It measures how much unpaid labour is squeezed from each unit of paid labour. It can be raised in two main ways: by lengthening the working day or intensifying work (absolute surplus value), or by cheapening the goods that workers consume, so that less of the day is needed to reproduce labour power (#relative_surplus_value). Technology is the main weapon of the second method. The organic composition of capital is: q = c / v Here a distinction matters. The technical composition is a physical ratio: how many tons of steel and how many machines per worker. The value composition is the ratio of the values of c and v. The #organic_composition_of_capital is the value composition insofar as it reflects changes in the technical composition. The distinction is not pedantic. If machines become physically more numerous per worker but each machine becomes much cheaper, the value composition may rise slowly, stagnate, or even fall for a period. This is the single most important escape hatch in the entire theory, and Section 7 returns to it. 3.4 The rate of profit The general #rate_of_profit is: r = s / (c + v) Divide numerator and denominator by v: r = (s/v) / ((c/v) + 1) = e / (q + 1) This deceptively small equation contains the whole argument of the paper. The profit rate rises when the rate of exploitation rises and falls when the organic composition rises. The law of the tendency of the rate of profit to fall states that, over the long run, the drive of competition compels capitals to raise labour productivity by adopting more and more means of production per worker; q rises; and although e also rises, it cannot rise without limit, because a working day has only twenty-four hours and a worker cannot be paid less than nothing. Therefore, eventually, the denominator grows faster than the numerator, and r declines. Marx himself insisted this is a tendency, not an iron fate. He devoted an entire chapter to the #counteracting_tendencies that slow, suspend or temporarily reverse it. Any honest use of the theory must give those counteracting forces their full weight, and Section 7 does so. 3.5 Why this framework is the right tool for AI Three properties of AI make Marx's framework unusually well suited to the problem. First, AI is capital-intensive at the frontier. Training a large model requires a scale of #fixed_capital that only a handful of firms can command, while the marginal cost of running a trained model is comparatively small. This is the classic profile of a rising technical composition. Second, AI directly targets #living_labour, and it does so in occupations that were previously considered safe from #machinery: writing, coding, analysis, translation, design, customer service. It therefore threatens to reduce v across a very wide front at once. Third, AI produces a commodity that is close to non-rival. A trained model can be copied at almost no cost, which means its value, in the sense of socially necessary labour time to reproduce it, collapses rapidly once the technique is known. This creates a permanent tension between the enormous cost of production and the near-zero cost of reproduction, which capital resolves only by legal and technical enclosure, that is, by converting profit into #rent. Section 9 develops this point. 4. Methodology 4.1 Research design This is a theoretical paper. Its method is conceptual analysis combined with formal modelling and a structured reading of secondary empirical evidence. It does not generate new primary data, and it does not claim to. Three methodological procedures are used. Immanent critique. The paper takes Marx's categories on their own terms and asks whether they can absorb a phenomenon they were not designed for. This is a standard procedure in theoretical political economy. A theory is strengthened not by being protected from hard cases but by being pushed into them. The hard case here is a machine that appears to perform cognitive labour. Comparative-static modelling. The paper builds a simple value model and varies the parameters that AI adoption is expected to change: the organic composition, the rate of surplus value, the turnover time of capital, and the depreciation rate of fixed capital. The purpose is not prediction but the identification of conditions: exactly how fast must #exploitation rise to offset a given rise in #capital_intensity? Numerical illustrations make the conditions concrete. Structured evidence review. Empirical claims made in Sections 8 and 10 are drawn from published peer-reviewed studies and monographs, selected on three criteria: recency, methodological transparency, and direct relevance to at least one parameter of the model. Where the evidence is contested, the contest is reported rather than resolved. 4.2 Definitions and scope "AI" in this paper means the family of statistical learning systems currently deployed at scale in production: supervised and self-supervised models, large language and multimodal models, recommendation and forecasting systems, computer vision, and the robotic systems in which such models are embedded. It does not mean hypothetical artificial general intelligence. Where the argument depends on the distinction, this is stated. "Automation" means the substitution of machine operation for human operation in a task previously performed by a worker. It is a task-level concept, not an occupation-level one, and this matters, because most occupations are bundles of tasks and it is unusual for AI to remove all of them. "Value" always means labour value in Marx's sense unless the text explicitly says market value or price. 4.3 Limits of the method Three limits should be stated at the outset rather than buried in the conclusion. First, labour values are not directly observable, so any empirical test must use money-value proxies with known imperfections. Second, the boundary between productive and unproductive labour, which affects every measured profit rate, is contested among Marxist economists themselves. Third, AI diffusion is young; the data series available today are too short to settle long-run questions. The paper therefore aims at conditional propositions, not verdicts. 5. Is Artificial Intelligence Constant Capital? 5.1 The classification problem The entire argument of the paper rests on one classification. If AI belongs to #constant_capital, then its spread reduces the base of value creation and pressures profitability. If AI belongs to variable capital, that is, if it is a new source of value rather than a transmitter of old value, then Marx's law dissolves and the theory must be rewritten. The intuitive case for treating AI as something new is easy to state. A hammer does nothing without a hand. A language model, by contrast, appears to do something: it drafts, summarises, decides, corrects. It looks less like a tool and more like a worker. If it behaves like a worker, why not treat it as one? 5.2 Four reasons the classification holds First, value creation is a social relation, not a technical performance. In Marx's theory, new value is created not because a physical operation is performed but because a specific social relation exists: a person who owns nothing but their capacity to work sells that capacity, for a definite time, to a person who owns the means of production. The surplus arises from the gap between the value of #labour_power and the value that labour power creates when it is consumed in production. A machine has no labour power to sell, because it has no separate existence from which to sell it. It is already owned. Its wage is its purchase price plus its running costs, and it cannot systematically produce more value than it cost, because the market for machines, like any other commodity market, drives the price of a machine toward its cost of production. Second, competition erases any temporary advantage. Suppose a firm buys an AI system that lets it produce at half the labour cost of its rivals. The firm's individual profit soars. But this is not new value created by the machine; it is a transfer of surplus value from other capitals, obtained because the firm sells above its individual value while below the prevailing social value. As soon as competitors adopt the same system, the socially necessary labour time for the product falls, the market price falls with it, and the extra profit disappears. What remains is a permanently higher ratio of machinery to workers. This mechanism, which Marx describes in his treatment of #relative_surplus_value, is precisely why individual firms adopt technology that harms the class of capitalists as a whole. Third, the model is dead labour in the most literal sense. A trained model is a compressed statistical representation of an enormous corpus of human-produced text, images, code, speech and behaviour, refined by human annotators and reinforced by human feedback. Pasquinelli (2023) makes this point with historical precision: machine intelligence is the codification of collective human knowledge and of the division of labour itself. Crawford (2021) makes it materially: the model runs on minerals extracted by miners, in buildings raised by construction workers, powered by electricity generated by other workers. The autonomous system is a dense knot of past and present human effort. Calling it #dead_labour is not a metaphor; it is an accounting statement. Fourth, the falsification condition is not met. The honest way to state a theory is to name the condition under which it would fail. AI would cease to be constant capital if an artificial agent were to exist as a legal person separate from its owner, reproduce itself and its own maintenance requirements out of a share of the product, and sell its capacity to work in a market where it could also refuse to sell. Only then would there be a gap between the value of its labour power and the value it creates, and only then would it produce surplus value. Dyer-Witheford, Kjosen and Steinhoff (2019) explore this scenario seriously and conclude that it would describe not capitalism with better machines but a different mode of production entirely. Nothing in current AI development approaches this condition. Present systems are commodities: owned, bought, depreciated and written off. 5.3 The special features that do matter Classifying AI as constant capital does not mean AI is just another machine. Four features distinguish it and complicate the analysis. Non-rivalry. Once a model exists, copying it costs almost nothing. Its reproduction value therefore tends toward zero, even though its original production cost was enormous. In value terms this is a machine whose value begins to evaporate the moment it exists. Extreme #moral_depreciation. Marx used this term for the value lost by machinery not through wear but through the appearance of better or cheaper machinery. In AI, moral depreciation is brutally fast: a frontier model can lose most of its competitive value within a year. Firms must therefore write off enormous #fixed_capital over very short lives, which raises the annual depreciation charge that has to be recovered from a shrinking base of #living_labour. Generality. Unlike a stamping press, a general-purpose learning system can be applied across many sectors at once. The rise in the organic composition it causes is therefore not confined to one industry; it propagates through the whole economy, including services that were previously labour-intensive. Dependence on continuing human input. Models degrade without fresh data, fresh annotation, fresh feedback and constant maintenance. The machine is not a one-time purchase but a subscription to a stream of ongoing human labour, much of it invisible. Section 8 pursues this point. 6. A Formal Treatment: Organic Composition and the Rate of Profit under AI Adoption 6.1 The baseline identity Recall the four expressions: Value of output: W = c + v + s Rate of surplus value: e = s / v Organic composition: q = c / v Rate of profit: r = s / (c + v) = e / (q + 1) If we allow for fixed capital that is used but not fully consumed in a period, the profit rate on the capital advanced is: r = s / (K + v) where K is the stock of constant capital advanced, while the c appearing in the value of output is only the portion transferred in the period, that is, depreciation plus materials. This distinction matters for AI, because the stock of capital advanced is enormous relative to the portion consumed in any single year, and because the depreciation portion is unusually high owing to moral depreciation. Turnover must also be included. If capital turns over n times per year, the annual rate of profit rises with n. Digital technologies genuinely accelerate turnover: orders, payments, logistics and delivery are faster. This is a real offsetting force and Section 7 returns to it. 6.2 Numerical illustration 1: the pure automation case Take a firm, and then by extension an economy, that begins with the following structure, in arbitrary value units: c = 80, v = 20, e = 100 per cent, therefore s = 20 q = 80 / 20 = 4 r = 20 / 100 = 20 per cent Now introduce AI. The firm replaces half its workforce with machines. Suppose it spends an additional 40 on constant capital and cuts variable capital to 10. Assume, generously, that the surviving workers are made to work harder, so the rate of surplus value rises from 100 per cent to 150 per cent. c = 120, v = 10, e = 150 per cent, therefore s = 15 q = 120 / 10 = 12 r = 15 / 130 = 11.5 per cent The rate of exploitation rose by half. The rate of profit still fell by almost half. This is the mechanism in its purest form. The mass of surplus value fell from 20 to 15, while the capital that must be valorised rose from 100 to 130. 6.3 Numerical illustration 2: how much extra exploitation would be needed? What rate of surplus value would be required to hold r at 20 per cent with the new capital structure? We need: s / 130 = 0.20, so s = 26, so e = 26 / 10 = 260 per cent. The rate of exploitation would have to rise from 100 per cent to 260 per cent. In plain terms: if a worker previously spent four hours of an eight-hour day reproducing their own wage and four hours producing surplus, they would now have to spend about 2.2 hours on their wage and 5.8 hours on surplus. That is not impossible. It is pursued through wage suppression, work intensification, longer effective hours and cheaper wage goods. But it becomes progressively harder. As v shrinks, each further rise in e yields less additional surplus value in absolute terms, because e is a ratio applied to an ever smaller base. This is the asymmetry at the heart of the law: there is no upper bound on q, but there is a hard upper bound on e. Even at infinite exploitation, where workers were paid nothing at all, the maximum surplus is limited by the total living labour actually performed. 6.4 The limit case Push the logic to its end. Imagine a fully automated firm with v = 0. Then s = 0, because surplus value comes only from unpaid living labour, and r = 0 / (c + 0) = 0. The firm produces an enormous mass of use values and not a single unit of new value. It could still make money, of course, by selling those use values at prices that capture surplus value produced elsewhere in the economy. But the economy as a whole cannot do this, because there is no elsewhere. This is the reductio that Marx's critics find absurd and his defenders find decisive. Carchedi and Roberts (2023) put it bluntly: a fully automated capitalism is a contradiction in terms, because capitalism is a system for the production of value, and value is produced by human labour alone. The limit case matters even though full automation will not occur. It shows the direction of the pressure. Every step toward it squeezes the ratio. 6.5 The Okishio objection and the response The most serious formal challenge to the law is the Okishio theorem, first advanced in 1961 and still the standard reason why mainstream and some Sraffian economists dismiss Marx's law. The theorem states that if a capitalist adopts a new technique only when it reduces costs at current prices, and if the real wage remains constant, then the new general rate of profit, computed after prices adjust, cannot be lower than the old one. In short: rational technical choice cannot by itself lower the profit rate; only rising wages can. Three responses matter. The temporal response. The theorem compares two simultaneous equilibria. It assumes that inputs and outputs are valued at the same set of prices. In real accumulation, capital is advanced at one moment and the product is sold at a later moment, by which time technical progress has already devalued the inputs. Once valuation is treated as temporal and sequential rather than simultaneous, the theorem's result no longer holds, and profit rates can fall even with a constant real wage. This is the core of the temporal single-system interpretation, defended in recent work by Carchedi and Roberts (2023). The fixed capital response. Okishio's original framework handles fixed capital awkwardly. When machinery is long-lived and subject to heavy moral depreciation, a technique that lowers unit costs for the innovating firm can still lower the general profit rate once the devaluation of the existing capital stock is counted. AI is precisely such a case, since it makes prior capital stocks obsolete at high speed. The empirical response. Even granting the theorem inside its own assumptions, those assumptions include a constant real wage, full price adjustment, no unused capacity, no credit, no rent and no monopoly. If long-run profit rates in fact decline, as Basu, Huato, Jauregui and Wasner (2023) report for the world economy across the period from 1960 to 2019, then the theorem tells us that at least one of its assumptions fails in reality. It does not tell us that the observed decline is impossible. 6.6 What the model predicts for AI The model yields four propositions that can, in principle, be tested. P1. AI diffusion raises the technical composition of capital across a broad range of sectors, including services. P2. Whether it raises the value composition depends on the price of computation. If the cost of a unit of AI capability falls faster than such units are accumulated, the value composition can remain flat for a period. P3. The mass of surplus value falls where AI displaces living labour, unless the rate of exploitation rises sharply among the remaining workers, or unless displaced workers are absorbed into other value-producing activity. P4. In the medium term the general rate of profit falls, but the distribution of profit becomes extremely unequal, because the firms that own the models capture surplus value produced elsewhere in the form of #rent. P4 is the proposition that best matches what we currently observe. Section 9 examines it. 7. Counteracting Tendencies in the Age of AI Marx listed the forces that offset the tendency. Every one of them is active today, several in intensified form. An analysis that ignores them is not Marxism but fatalism. 7.1 Cheapening the elements of constant capital This is the most powerful offset and the one most relevant to AI. Technical progress does not only add machines; it makes machines cheaper. The cost of a unit of computation has fallen by orders of magnitude over decades. The cost of training a model of a given capability falls rapidly year by year as algorithms improve, as hardware improves, and as techniques such as distillation and quantisation spread. The consequence is that the value composition may rise far more slowly than the technical composition. A firm may deploy a hundred times more computing power per worker than a decade ago while spending only three or four times as much on it in value terms. This is exactly why the empirical record of profitability shows decline punctuated by long recoveries rather than a smooth collapse. But note the limit. Cheapening constant capital slows the rise in q; it does not raise the mass of surplus value. And the cheapening is itself achieved by displacing living labour in the capital goods sector, which reduces the value produced there. The offset works, but it works by feeding the very process it offsets. 7.2 Raising the rate of exploitation AI is a formidable instrument for raising the rate of surplus value among the workers who remain. Intensification. Algorithmic management sets the pace, tracks the worker, scores performance and removes the small pockets of slack that human supervision leaves. Delfanti (2021) shows how warehouse robotics do not simply replace pickers but restructure their movements so that more work is extracted per hour. Altenried (2022) generalises this to the digital factory across sectors. Extension. Always-available communication tools stretch the working day into the evening and the weekend without stretching pay. Cheapening of wage goods. If AI lowers the cost of food, transport, housing services, education and healthcare, then the value of #labour_power falls and relative surplus value rises without any change in the length or intensity of the working day. This is the classic route, and it is genuinely available. Deskilling and wage compression. When a task is codified into a model, the skill premium attached to it collapses. Workers who once commanded high wages for judgement now supervise a machine that performs the judgement, and their bargaining position erodes. Acemoglu and Restrepo (2022) document the wage consequences of exactly this dynamic in earlier waves of automation. 7.3 The relative surplus population Automation creates what Marx called the industrial reserve army: a pool of unemployed and underemployed people whose existence disciplines those in work. AI extends the reserve army into occupations that previously felt secure, including clerical, administrative, creative and paralegal work. A larger reserve army lowers wages, which raises e, which raises r. Benanav (2020) argues persuasively that the labour-market symptom of the last half-century is not mass technological unemployment but chronic underemployment and the growth of low-productivity service work, which is what a growing reserve army actually looks like. 7.4 Foreign trade and the global relocation of labour If a firm imports cheap inputs produced by low-wage labour abroad, the value of its constant capital falls and the value of its workers' consumption basket falls. Both raise the profit rate at home. Much of the apparent capital-intensity of rich-country production is possible only because the labour-intensive parts of the chain have been relocated. This is not a marginal detail; it is an organising fact of the world economy. AI supply chains follow the same pattern: chips and devices assembled where wages are low, data labelled where wages are lower still, models sold where incomes are highest. 7.5 Acceleration of turnover Digital coordination compresses the time between the advance of capital and its return. Since the annual rate of profit rises with the number of turnovers, faster circulation raises measured profitability even when the profit rate per turnover is unchanged. This is a genuine gain and one of the clearest real contributions of information technology to capital. 7.6 Interest, rent and financial claims Some capital is not invested to produce surplus value but to claim a share of it. The growth of interest-bearing capital, of platform rents and of intellectual property income moves surplus value around without producing it. For the individual owner this looks like profit. For the system, it is redistribution. This point is central and is treated separately in Section 9. 7.7 Devaluation and crisis as the ultimate corrective Marx's darkest counteracting tendency is crisis itself. When profitability falls far enough, investment stops, firms fail, and capital is destroyed or written down. The denominator of the profit rate shrinks. Assets are bought cheaply by the survivors, and accumulation restarts on a lower base. In this reading, the periodic destruction of capital value is not a malfunction of the system but its way of restoring the conditions of profitability. Applied to AI, the implication is uncomfortable: a large write-off of AI-related #fixed_capital would not refute the theory. It would confirm it. 8. The Hidden Living Labour Behind Artificial Intelligence 8.1 The illusion of the labour-free machine The most persistent myth in the AI debate is that AI systems are labour-free. They are not. They are labour-dense, but the labour is arranged so that it is difficult to see. It happens in other countries, in other legal categories, under other names, and it is deliberately kept out of the product description. Four kinds of #living_labour sustain every deployed model. Data work. Models learn from labelled examples. Someone must draw the boxes around pedestrians, mark the tumours, rank the answers, write the preferred response, and flag the toxic output. Muldoon, Graham and Cant (2024) document this workforce in detail across several countries, describing pay rates, precarity and psychological cost. Gray and Suri (2019) named the phenomenon #ghost_work: labour that is essential to the product, invisible in the interface, and structurally denied the status of employment. In value terms this is straightforwardly productive labour: it is bought with variable capital, it produces a commodity, and it produces surplus value. It is simply cheap. Content moderation. Systems that must not produce harmful output are made safe by workers who read and view the harmful material so that the model can be trained away from it. This is #exploitation in the technical Marxist sense and, frequently, in the everyday moral sense as well. Logistics and physical fulfilment. The visible face of platform capitalism is an app; the invisible body is a warehouse, a van and a kitchen. Delfanti (2021) shows that the most automated warehouses in the world employ enormous numbers of people, working faster than before. Infrastructure labour. Data centres are built by construction workers, wired by electricians, cooled by systems maintained by technicians, and fed by electricity generated by power-plant workers. The chips are fabricated in plants that cost tens of billions and employ thousands. The minerals are mined, often in conditions that would be recognisable to a reader of the nineteenth-century factory reports, as Crawford (2021) documents. 8.2 Why this matters for value theory If AI depended on no living labour at all, the theory would predict a collapse in the mass of surplus value. It predicts no such thing, because AI does depend on living labour, and a great deal of it. What has changed is the composition and geography of that labour, not its existence. Three consequences follow. First, the aggregate mass of surplus value falls more slowly than a naive automation story implies, because a large volume of poorly paid, highly exploited labour has been added to the world working class even as better-paid labour is displaced in the rich countries. A high rate of surplus value on a large mass of cheap labour can sustain profitability for a long time. Second, the appearance of a knowledge economy conceals a reality of expanded manual and clerical drudgery. Jason E. Smith (2020) makes this argument forcefully: the striking feature of contemporary capitalism is not the vanishing of work but the proliferation of low-productivity service work in which #automation has barely advanced. Third, the political economy of AI cannot be separated from questions of global inequality. Value produced in one part of the world is realised as profit in another. The pattern is old. Only the technology is new. 8.3 Users as unpaid contributors A harder question concerns users. When people write posts, upload images, correct autocomplete or click on results, they generate the raw material for models. Are they producing value? The safest answer within Marx's framework is no, because they do not sell #labour_power and they are not organised in a labour process directed by capital. What they do produce is a free raw material, a gift of use value that lowers the cost of #constant_capital for the platform. That is significant: it is a subsidy, not a source of surplus value. Fuchs (2022) argues the stronger position, that user activity is directly value-producing digital labour. This paper takes the weaker and, in its view, more defensible position, while acknowledging the debate remains open. 9. Rent, Intellectual Monopoly, and the Redistribution of Surplus Value 9.1 The puzzle of extraordinary profits If the theory predicts falling profitability, how do we explain the extraordinary profits of the largest technology firms? This is the objection every student raises, and it deserves a direct answer. The answer is that a high profit for a firm is not the same thing as a high rate of profit for the system. A firm can be enormously profitable by capturing surplus value produced elsewhere. Marx analysed exactly this under the heading of #rent: a payment made to the owner of a scarce and monopolised condition of production, drawn out of the surplus value produced in the economy at large. 9.2 The mechanisms of technological rent Four monopolised conditions produce technological rent today. Intellectual property. Patents, copyrights, licences and trade secrets convert freely copyable information into an artificially scarce commodity. Without enclosure, a model with near-zero reproduction cost would have near-zero price. Rikap (2021) shows how a small number of firms have built systems of #intellectual_monopoly, capturing the results of research performed across universities, public laboratories and smaller firms. Data. Access to large proprietary datasets is a barrier to entry that cannot be overcome with money alone, because the data are a by-product of a user base that competitors do not have. Compute. The physical capacity to train frontier models is concentrated in a handful of firms and a handful of chip designers and foundries. This is a genuine scarcity, and it commands a genuine rent. Network effects and default position. Being the default assistant, the default search box or the default cloud provider creates a self-reinforcing advantage that has nothing to do with the labour content of the service. 9.3 Rent is not value creation The crucial theoretical point, and the one that resolves the puzzle, is this. Rent redistributes surplus value; it does not create it. When a firm pays a large licence fee for an AI service, that payment comes out of the surplus value produced by the workers of the paying firm. The technology owner's gain is the technology user's loss. Across the whole economy the sum is zero, minus the real labour actually performed in producing and maintaining the system. This produces exactly the picture we observe: spectacular concentration of profit at the top, thin margins in the middle, and an average rate of profit that stagnates or declines. Durand (2024) and Varoufakis (2023) both describe this configuration and both reach for the vocabulary of feudalism to name it. This paper resists that vocabulary. Rent is not feudal; rent is a normal, well-theorised category inside capitalism, and Marx devoted a large part of the third volume of Capital to it. What we are seeing is not a new mode of production but capitalism in which the rentier moment has become dominant, which is precisely what one would expect in a period of weak underlying profitability. Capital that cannot find profitable productive investment flows into claims on other people's surplus value. 9.4 Implications If this analysis is right, three things follow. First, measured aggregate profitability will look better than underlying productive profitability, because rents inflate the numerator for some firms while the productive base erodes. Second, investment in productive capacity will remain weak relative to the scale of profits, because rent-seeking is more attractive than production. This matches the widely observed pattern of high profits accompanied by mediocre investment. Third, the system becomes more fragile. A structure in which a few firms capture surplus value produced by everyone else, financed by heavy borrowing against expected future rents, is vulnerable to any disappointment in those expectations. 10. Empirical Signals The theory cannot be tested directly, but four bodies of evidence are consistent or inconsistent with it, and honesty requires reporting both. 10.1 The labour share A rising organic composition should be accompanied, other things equal, by a falling share of wages in output. The labour share in most advanced economies has indeed declined since around 1980. Autor, Dorn, Katz, Patterson and Van Reenen (2020) attribute much of this fall to the rise of superstar firms: industries have become more concentrated, and the leading firms have unusually low labour shares. This is consistent with both the rising-composition story and the rent story, and it does not by itself discriminate between them. Acemoglu and Restrepo (2022) attribute a large share of wage stagnation for less-educated workers to task displacement, which is closer to the mechanism proposed here. 10.2 Profit rates Basu, Huato, Jauregui and Wasner (2023) reconstruct profit rates for the world economy across six decades and find a long-run downward trend, with a partial recovery in some periods. Tsoulfidis and Tsaliki (2019) find similar long-wave patterns and connect them to the movement of capital intensity. These findings are not decisive. Measurement of the capital stock, of depreciation and of the productive-unproductive boundary all involve choices that affect the result. But the broad direction is consistent with the theory, and it is not what the Okishio theorem would lead one to expect. 10.3 The productivity paradox If AI were as transformative as its promoters claim, aggregate productivity growth should be accelerating. It is not, or at least not yet. Acemoglu (2025) estimates the likely macroeconomic gains from generative AI over a decade and finds them modest. This is a paradox for the enthusiasts, but it is not a paradox for the theory advanced here. A technology can raise physical productivity enormously in specific tasks while contributing little to aggregate value, because value is not physical output. Indeed, an enormous rise in physical output accompanied by stagnant value output is exactly what the labour theory of value predicts: more use values, embodying less #socially_necessary_labour_time each. Micro-level studies do find real effects. Noy and Zhang (2023) report substantial time savings in professional writing tasks in a controlled experiment. Brynjolfsson, Li and Raymond (2025) find productivity gains among customer-support agents, concentrated among the least experienced workers, which suggests the technology works partly by codifying the tacit knowledge of the most skilled and redistributing it downward. Eloundou, Manning, Mishkin and Rock (2024) estimate that a large fraction of occupations have at least some tasks exposed to language models. Note what all three findings imply in Marxist terms: faster task completion means the same wage buys more labour, that is, a higher rate of #exploitation. They do not imply that the model creates value. 10.4 The material and energy cost Constant capital is not only chips. The energy footprint of AI computation has grown rapidly, and de Vries (2023) provides an early quantification of the electricity demand implied by large-scale deployment. Energy is a component of c. Rising energy costs raise the denominator of the profit rate and therefore reinforce, rather than offset, the tendency under discussion. This is a strand of the argument that deserves far more attention than it currently receives. 10.5 What the evidence does not show Intellectual honesty requires stating the negative. There is, at present, no direct empirical demonstration that AI adoption has raised the value composition of capital economy-wide, because the necessary data do not yet exist in usable form. There is no consensus profit-rate series. And the periods in which profitability recovered, notably from the mid-1980s to the mid-2000s in several economies, show that #counteracting_tendencies can dominate for decades. Anyone claiming that AI must produce an imminent crisis is going beyond the evidence. The claim defended here is narrower: AI intensifies a structural pressure, and the pressure has to be resolved somehow, either by raising exploitation, by expanding rent, or by devaluing capital in a crisis. 11. Discussion: Three Scenarios Bringing the argument together, three scenarios can be distinguished. They are not predictions but conditional paths, each defined by which parameter in the model dominates. 11.1 Scenario A: the exploitation path Here the rise in q is offset by an aggressive rise in e. Wages stagnate, work intensifies, working hours effectively lengthen, wage goods cheapen, and a growing global reserve army of labour keeps the price of #labour_power low. The profit rate holds up. Physical output and consumption rise. Inequality widens sharply, and the social experience is one of insecurity amid abundance. This is essentially a continuation of the last four decades, with AI as the newest instrument. It is, in the author's assessment, the most likely medium-term path. 11.2 Scenario B: the rent path Here productive profitability continues to erode, but a small number of firms sustain extraordinary returns by monopolising models, data and compute. Investment is channelled into assets that yield claims rather than into productive capacity. Financial valuations run far ahead of the underlying surplus value available to service them. The system does not collapse, but it becomes brittle, sluggish and politically explosive, because the visible gap between the rewards of the owners and the rewards of the workers becomes impossible to justify. A large devaluation of AI-related #fixed_capital is a live possibility on this path, and it would be a textbook case of Marx's crisis mechanism at work rather than evidence against it. 11.3 Scenario C: the transformation path Here the contradiction is resolved by changing the social relations rather than by intensifying them. If machines can produce the necessities of life with a fraction of the labour once required, then the problem of distributing those necessities is not technical but political. The productive capacity that makes profit impossible is the same capacity that would make free time possible. This is the point at which analysis ends and politics begins, and it is the point Marx himself reached in the Fragment on Machines. Whether one calls the outcome #post_capitalism, socialism, or something else, the analytical claim is modest: a system that requires human labour to produce value cannot survive the elimination of human labour, so the elimination of human labour requires a system that does not require it. 11.4 Synthesis The three scenarios are not mutually exclusive. Elements of A and B are visible simultaneously right now. The theoretical conclusion of this paper can be stated in a single sentence: AI does not falsify the labour theory of value; it makes the theory's central contradiction more acute, and it shifts the burden of maintaining profitability from the production of surplus value to its extraction and redistribution. 12. Implications 12.1 For students and researchers Students working in computer science, information systems and management should take one lesson from this analysis above all others: technology is never neutral with respect to social relations, and the question "who benefits?" is an engineering question as much as a political one. Those working in economics should notice that a theory declared dead many times keeps producing predictions that the data do not obviously refute, which is more than can be said for several fashionable alternatives. There is also a methodological lesson. The debate about AI and work is dominated by two failure modes: uncritical hype, and uncritical doom. Both share the assumption that the technology is the agent and society is the object. Marx's framework reverses this. Machines do not decide anything. Owners decide, under the compulsion of competition. That is where analysis should look. 12.2 For policy Four implications follow, stated without partisanship. Taxation. If a large share of technology income is #rent rather than productive profit, then taxing it does not discourage production in the way that taxing productive profit might. The classical economists understood this about land rent, and the argument transfers. Working time. If productivity per hour rises while employment falls, the arithmetic solution is to reduce the working week rather than to reduce the number of workers. This is a distributive decision, not a technical one. Income support. Proposals for #universal_basic_income are a rational response to a value system that cannot employ everyone, but they do not address who owns the machines. On the analysis presented here, income support without ownership change transfers a share of surplus value without altering its source, which limits how far it can go. Data and compute governance. If monopolised data and compute are the basis of technological rent, then public or collective provision of compute and data infrastructure attacks the rent at its root. This is a live policy debate in many countries. 12.3 For the labour movement If AI raises the rate of #exploitation among remaining workers rather than simply removing them, then the site of struggle is not only "save my job" but "control the pace, the surveillance and the terms". The organising literature reviewed here, particularly Delfanti (2021) and Altenried (2022), suggests that the most effective points of leverage are logistical chokepoints and the annotation and moderation workforces, precisely because these are the human dependencies that the automated system cannot do without. 13. Limitations This paper has clear limits. It is theoretical, and it does not test its own propositions with original data. The four propositions in Section 6.6 are offered as a research agenda, not as findings. Its measurement claims depend on secondary literature whose methods are contested. Different treatments of unproductive labour, of depreciation, and of the valuation of the capital stock generate materially different profit-rate series. It takes a position in an unresolved internal debate about whether user activity constitutes value-producing labour, and readers who take the opposite view will reach different quantitative conclusions. It is written at a moment of intense uncertainty about AI capabilities. If systems were to acquire genuine autonomy and legal personality, the falsification condition set out in Section 5.2 would be met, and the theory as applied here would require substantial revision. Finally, the paper concentrates on production and pays comparatively little attention to realisation, that is, to whether the mass of commodities produced can actually be sold. A complete crisis theory must integrate both, and this one does not. 14. Conclusion The question this paper set out to answer was whether the rise of #artificial_intelligence destroys Marx's #labour_theory_of_value or vindicates it. The answer, on the evidence and argument assembled here, is neither a simple yes nor a simple no, but something more interesting. AI is #constant_capital. It is #dead_labour in an unusually pure form: the accumulated, codified, statistically compressed knowledge and activity of millions of people, owned by a few and set to work against the many. It transfers value; it does not create it. Because it is constant capital, its spread raises the #organic_composition_of_capital and squeezes the base of #living_labour from which #surplus_value is drawn. Everything else being equal, this depresses the general #rate_of_profit, exactly as Marx's law predicts. But everything else is not equal, and this is where the analysis becomes genuinely useful rather than merely dramatic. The cheapening of computation slows the rise in the value composition. The global relocation of labour and the addition of hundreds of millions of low-paid workers to the world labour force raise the mass of surplus value. Algorithmic management raises the rate of #exploitation among the workers who remain. Faster turnover raises the annual return. And, above all, the monopolisation of models, data and compute allows a small number of firms to capture as #rent the surplus value produced by everyone else, which makes the system look enormously profitable at the top while it stagnates underneath. The picture that emerges is not a machine-driven apocalypse and not a machine-driven utopia. It is a familiar capitalism, under new technical conditions, in which the pressures on profitability are managed by intensifying work, relocating it, hiding it, and taxing it through rent. The tendency of the rate of profit to fall is not a prophecy of collapse. It is a description of a permanent pressure and of the increasingly costly measures required to contain it. There is a final point, and it is the one worth leaving with a student reader. If human labour is the only source of value, then a society that eliminates human labour cannot be organised around value. The technical achievement of removing drudgery from human life is, on Marx's own account, an achievement worth having. The problem is not the machine. The problem is that under present arrangements the machine's productivity appears not as free time but as unemployment, not as abundance but as insecurity. That is a fact about property, not about processors. The #labour_theory_of_value is useful today precisely because it makes that distinction impossible to miss. References Acemoglu, D. (2025). The simple macroeconomics of AI. Economic Policy, 40(121), 13 to 58. Acemoglu, D., and Johnson, S. (2023). Power and Progress: Our Thousand-Year Struggle Over Technology and Prosperity. New York: PublicAffairs. Acemoglu, D., and Restrepo, P. (2020). Robots and jobs: evidence from US labor markets. Journal of Political Economy, 128(6), 2188 to 2244. Acemoglu, D., and Restrepo, P. (2022). Tasks, automation, and the rise in US wage inequality. Econometrica, 90(5), 1973 to 2016. Altenried, M. (2022). The Digital Factory: The Human Labor of Automation. Chicago: University of Chicago Press. Autor, D., Dorn, D., Katz, L. F., Patterson, C., and Van Reenen, J. (2020). The fall of the labor share and the rise of superstar firms. Quarterly Journal of Economics, 135(2), 645 to 709. Basu, D., Huato, J., Jauregui, J. L., and Wasner, E. (2023). World profit rates, 1960 to 2019. Review of Political Economy, 35(1), 149 to 176. Benanav, A. (2020). Automation and the Future of Work. London: Verso. Bender, E. M., and Hanna, A. (2025). The AI Con: How to Fight Big Tech's Hype and Create the Future We Want. New York: Harper. Brynjolfsson, E. (2022). The Turing trap: the promise and peril of human-like artificial intelligence. Daedalus, 151(2), 272 to 287. Brynjolfsson, E., Li, D., and Raymond, L. (2025). Generative AI at work. Quarterly Journal of Economics, 140(2), 889 to 942. Carchedi, G., and Roberts, M. (2023). Capitalism in the 21st Century: Through the Prism of Value. London: Pluto Press. Crawford, K. (2021). Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence. New Haven: Yale University Press. de Vries, A. (2023). The growing energy footprint of artificial intelligence. Joule, 7(10), 2191 to 2194. Delfanti, A. (2021). The Warehouse: Workers and Robots at Amazon. London: Pluto Press. Durand, C. (2024). How Silicon Valley Unleashed Techno-feudalism: The Making of the Digital Economy. London: Verso. Dyer-Witheford, N., Kjosen, A. M., and Steinhoff, J. (2019). Inhuman Power: Artificial Intelligence and the Future of Capitalism. London: Pluto Press. Eloundou, T., Manning, S., Mishkin, P., and Rock, D. (2024). GPTs are GPTs: labor market impact potential of large language models. Science, 384(6702), 1306 to 1308. Fuchs, C. (2022). Digital Capitalism: Media, Communication and Society, Volume Three. London: Routledge. Gray, M. L., and Suri, S. (2019). Ghost Work: How to Stop Silicon Valley from Building a New Global Underclass. Boston: Houghton Mifflin Harcourt. Heinrich, M. (2021). How to Read Marx's Capital: Commentary and Explanations on the Beginning Chapters. New York: Monthly Review Press. Moseley, F. (2023). Marx's Theory of Value in Chapter 1 of Capital: A Critique of Heinrich's Value-Form Interpretation. Chicago: Haymarket Books. Muldoon, J., Graham, M., and Cant, C. (2024). Feeding the Machine: The Hidden Human Labour Powering AI. Edinburgh: Canongate. Mueller, G. (2021). Breaking Things at Work: The Luddites Are Right About Why You Hate Your Job. London: Verso. Noy, S., and Zhang, W. (2023). Experimental evidence on the productivity effects of generative artificial intelligence. Science, 381(6654), 187 to 192. Pasquinelli, M. (2023). The Eye of the Master: A Social History of Artificial Intelligence. London: Verso. Rikap, C. (2021). Capitalism, Power and Innovation: Intellectual Monopoly Capitalism Uncovered. London: Routledge. Sadowski, J. (2025). The Mechanic and the Luddite: A Ruthless Criticism of Technology and Capitalism. Oakland: University of California Press. Smith, J. E. (2020). Smart Machines and Service Work: Automation in an Age of Stagnation. London: Reaktion Books. Steinhoff, J. (2021). Automation and Autonomy: Labour, Capital and Machines in the Artificial Intelligence Industry. Cham: Palgrave Macmillan. Tsoulfidis, L., and Tsaliki, P. (2019). Classical Political Economics and Modern Capitalism: Theories of Value, Competition, Trade and Long Cycles. Cham: Springer. Varoufakis, Y. (2023). Technofeudalism: What Killed Capitalism. London: The Bodley Head. Woodcock, J. (2021). The Fight Against Platform Capitalism: An Inquiry into the Global Struggles of the Gig Economy. London: University of Westminster Press. Hashtags #Marxist_Economics #AI_and_Labour #Political_Economy_of_Technology #Value_Theory #Automation_Studies #Digital_Capitalism #Future_of_Work #Machine_Learning_Economics #Capital_Accumulation #Technological_Unemployment #Critical_AI_Studies #Rate_of_Profit_Debate #Labour_Process_Theory #Platform_Economy #Economics_Research

This paper argues that the urban transformations grouped under the label of gentrification are best understood not as a market curiosity or a matter of consumer taste, but as a continuing form of #primitive_accumulation. Karl Marx devoted the closing part of the first volume of Capital to the expropriation of the agricultural population from the land, describing how peasants were driven off common fields by law, force and administrative decree so that land could be turned into private capital. This paper takes that historical account seriously as an analytical model and applies it to the contemporary city. It treats #urban_gentrification, the bulk purchase of homes by financial firms, and the fencing of shared urban space as three faces of a single process of #spatial_enclosure. The study is conceptual and comparative. It synthesises recent empirical scholarship on housing financialisation, state led regeneration, green gentrification, tourism led displacement and public space privatisation, and reads that evidence through the categories of expropriation, enclosure and rent. Four findings emerge. First, contemporary displacement depends on the same combination of legal instruments and physical force that characterised the historical enclosures, so the term primitive is misleading if it is taken to mean early or finished. Second, the arrival of asset managers and #real_estate_investment_trusts as landlords converts dwellings from shelter into income streams and makes #eviction a routine management technique rather than an exception. Third, the closure and commercialisation of streets, parks, squares and estates removes the shared spaces that allow poor households to survive on low incomes, and so functions as an urban equivalent of the loss of grazing, gleaning and firewood rights. Fourth, the value released by these processes flows overwhelmingly to #rentier_capital rather than to production, employment or public revenue. The paper concludes that planning practice which frames displacement as an unfortunate side effect of investment misreads the mechanism, and that policies of #decommodification, tenant power and commons protection address the cause rather than the symptom. Keywords: primitive accumulation, gentrification, urban enclosure, corporate landlords, housing financialisation, public space, displacement, rent theory 1. Introduction The modern city is often described as a place where old buildings are repaired, poor areas improve, and neighbourhoods slowly change character as new residents move in. Told this way, the story is about #land_value and taste. Warehouses become lofts. Corner shops become coffee shops. Rents rise because more people want to live in the area. The residents who leave are said to have moved on, as if the departure were a private decision made in the privacy of a kitchen. This paper argues that the story is wrong in an important way. It confuses the surface appearance of a process with the process itself. What looks like a slow drift of taste is in fact a transfer of property, of security, and of the right to remain. It is organised, it is planned, it produces winners who are identifiable, and it produces losers who are also identifiable. When a #public_housing estate is demolished and rebuilt with a minority of below market units, this is not a change of fashion. When a #private_equity fund buys eleven thousand family houses in a single metropolitan region and raises rents across the portfolio at the same time, this is not a change of fashion either. Something older and harder is at work, and Marx gave it a name. In the final part of the first volume of Capital, Marx described what he called the so called primitive accumulation. The English peasantry, he wrote, did not become a working class because they saved money or because they chose wage labour. They became a working class because they were separated from the land. Common fields were fenced. Grazing rights were extinguished by statute. Cottages were pulled down. Whole villages were cleared to make room for sheep and later for game. Marx insisted that this history was written in letters of blood and fire, and he mocked the polite economic tale in which capital originated in the thrift of the industrious. The separation of people from their means of subsistence, he argued, is not a distant prelude to capitalism. It is the condition on which capitalism rests, and it must be constantly renewed. The claim of this paper is that the twenty first century city is one of the main places where that renewal now happens. The #enclosure_of_the_commons has not ended. It has moved from the open field to the housing estate, from the manor to the investment fund, and from the parish register to the planning application. The instruments have changed. Sheriffs have become bailiffs and asset management platforms. Acts of parliament have become regeneration frameworks, #compulsory_purchase orders, zoning variances and business improvement districts. But the underlying operation is recognisable: people who hold a customary, informal or weakly protected claim over space are separated from that space by a mixture of law and force, and the space is then reorganised so that it yields a stream of income to an owner who did not previously hold it. This argument sits inside a wider tradition. David Harvey reworked Marx's category into the idea of #accumulation_by_dispossession, arguing that when capital cannot easily expand through production it turns instead to seizing assets that already exist, especially public assets and common goods (Harvey, 2003). Urban scholars have applied that idea to #demolition, to slum clearance, to informal settlement eviction and to the sale of municipal land (Ibrahim, Abubakari, Cobbinah and Kuuire, 2024). Others have traced the way planning itself has functioned historically as an anti commons instrument, a discipline that arose in part to break up collective uses of land and to fix populations in governable patterns (Sevilla-Buitrago, 2022). The contribution attempted here is to bring three strands that are often studied separately into one analytical frame. The first strand is #urban_gentrification in the classic sense: the reinvestment of capital in devalued central areas and the replacement of lower income residents by higher income ones. The second strand is the corporate acquisition of housing, which includes the purchase of foreclosed homes after the financial crisis, the rise of listed residential landlords, and the entry of asset managers into rental markets (Taylor and Aalbers, 2024; Christophers, 2023). The third strand is the enclosure of shared urban space, which includes the sale of parks and plazas to private management, the conversion of streets into consumption zones, the greening of waterfronts in ways that raise rents, and the design of benches and doorways to prevent poor people from resting in them. These three strands are usually treated as different literatures. Gentrification research talks about neighbourhood change and culture. Housing finance research talks about capital markets and yields. Public space research talks about design, policing and citizenship. This separation is convenient but analytically weak. It hides the fact that the three processes have the same function. Each one takes something that was accessible without payment, or accessible at a payment below its potential market rent, and converts it into a source of #ground_rent. Each one requires that the previous users be removed or made to pay. Each one is defended in public with the language of improvement, safety, sustainability and regeneration. Marx's chapters on enclosure are useful precisely because they show this triple structure: expropriation, legalisation and moral justification. 1.1 Research problem The problem addressed by this paper can be stated simply. Urban policy in most countries treats #displacement as an unintended consequence of otherwise desirable investment. Planning documents speak of managing the risk of displacement, of mitigating impacts, of ensuring that regeneration is inclusive. This framing assumes that investment and displacement are separable, that one can be kept while the other is reduced. If the analysis offered here is correct, that assumption is false in a large class of cases. Where the profitability of an urban project depends on capturing the gap between current rent and potential rent, displacement is not a side effect of the project. It is the mechanism by which the project makes money. A policy that seeks investment while opposing displacement is, in those cases, asking for a fire that does not burn. 1.2 Research questions The paper is organised around four questions. In what precise ways do contemporary urban processes reproduce the structure of enclosure that Marx described, and in what ways do they differ? How do corporate and institutional owners of housing change the character of dispossession compared with small landlords and owner occupiers? What role does the privatisation of shared urban space play in the reproduction of #rentier_capital, and how does it affect the survival strategies of low income households? What follows for planning practice and housing policy if displacement is understood as a mechanism of accumulation rather than as a market accident? 1.3 Contribution The paper makes three contributions. Analytically, it offers an integrated reading of gentrification, corporate landlordism and public space privatisation as one process of enclosure, rather than three adjacent problems. Empirically, it synthesises evidence published mainly in the last five years, a period in which the ownership of urban housing changed faster than at any time since the postwar decades. Practically, it argues that the standard toolkit of affordable housing quotas and community consultation is poorly matched to the mechanism it claims to address, and it sets out what a #counter_enclosure agenda would involve. 1.4 Structure Section 2 sets out the theoretical framework, returning to Marx's account of expropriation and to the debate about whether primitive accumulation is a historical stage or a permanent feature. Section 3 reviews recent literature. Section 4 explains the method. Section 5 presents the analysis in four parts: gentrification as enclosure, corporate housing acquisition, the privatisation of public space, and the legal machinery that binds them. Section 6 discusses objections and counter arguments. Section 7 develops policy implications. Section 8 notes limitations. Section 9 concludes. 2. Theoretical Framework 2.1 What Marx actually said It is worth being careful about the source text, because the phrase primitive accumulation is often used loosely. Marx introduced the term in order to attack it. Classical political economy told a story in which some people worked hard and saved, while others were lazy and spent, so that in time the first group owned the means of production and the second group had nothing to sell but their labour. Marx called this an insipid childishness. Against it he set a historical account of how the English peasantry lost access to land. Three features of that account matter here. The first is that the process was not economic in the narrow sense. It did not happen through voluntary exchange in a market. It happened through the fencing of common land, the forcible clearing of estates, the destruction of dwellings, and a series of statutes that turned customary rights into trespass. Force preceded the market and made the market possible. This is why the concept cannot be reduced to competition. Wherever the state uses law, police or planning powers to break an existing claim on space so that a new owner can profit from it, the structure is the same. The second feature is that expropriation was accompanied by a legal fiction of legitimacy. The enclosures were not lawless. They were carried out through parliamentary acts, surveys, commissions and compensation schemes. The dispossessed were told that the process was orderly, that it was improving, that it raised productivity. Marx's irony was directed at exactly this: the theft was legal, and the law was made by the beneficiaries. Modern #urban_regeneration works the same way. It proceeds through consultation, statutory notice, viability assessment and compensation at market value, and it is precisely this procedural correctness that makes it hard to contest. The third feature is that the process created a population with no means of subsistence other than wages. Once the commons was gone, the peasant could not gather fuel, graze an animal, glean after harvest, or rest on land that was not owned. Subsistence became fully dependent on money. The importance of this point for urban analysis is often missed. Common resources are not merely pleasant. They are the difference between poverty and destitution. A city with free water fountains, public toilets, unpoliced benches, cheap street markets, rent controlled tenancies and open parks is a city in which it is possible to live on a very low income. Remove those things and the same nominal income becomes insufficient. This is why the #public_space_privatization discussed in Section 5.3 belongs in the same analysis as rent increases. Both raise the money price of staying alive in the city. 2.2 From primitive accumulation to accumulation by dispossession Marx's own text is ambiguous about time. In places he writes as if expropriation belonged to the prehistory of capital. In other places he treats colonial plunder, the slave trade and state debt as ongoing sources of the same kind of gain. Later scholars pulled on this thread. The core claim, developed across the twentieth century and consolidated by Harvey (2003), is that expropriation is not a stage that capitalism passes through and leaves behind. It is a permanent function that becomes more prominent when profits from production are hard to find. Harvey's argument runs roughly as follows. Capital that cannot be invested profitably in making things will look for other outlets. One outlet is the #built_environment, which absorbs enormous sums and produces long lived assets. Another outlet is the seizure of assets that already exist and can be bought cheaply, especially assets that were previously held in common or by the state. Privatisation of utilities, of pensions, of water, of land, and of housing are examples. The result is a form of growth that does not create new value so much as move existing value from one pocket to another. Harvey called this #accumulation_by_dispossession, and he treated the era of #neoliberal_urbanism as one in which it became the leading edge rather than a residual practice. This framing has been criticised, and the criticism is useful. Some argue that it stretches the concept until it covers any transfer that the analyst dislikes. Others argue that it is too tied to a Western reading and does not capture the political contestation of dispossession in the global South, where informality, customary tenure and postcolonial state building change the picture. The response taken here is to keep the concept narrow. Three tests are applied throughout this paper. A process is treated as enclosure only if all three are met. Test one: separation. A group loses access to a space or resource that it previously used, whether that use rested on ownership, tenancy, custom, informality or simple tolerance. Test two: extra economic force. The separation depends on something other than voluntary exchange: eviction, demolition, compulsory purchase, criminalisation, design that excludes, licensing that excludes, or administrative reclassification. Test three: conversion into a revenue stream. After separation, the space yields income to an owner in the form of rent, fees, higher property values or licensing revenue. Where all three hold, we are looking at #spatial_enclosure. Where only rising prices occur, without separation and without force, we are looking at something else, and the vocabulary of enclosure should not be used. This discipline matters, because the strength of the argument depends on not applying it everywhere. 2.3 Rent theory and the rent gap The bridge between Marx's chapters and contemporary urban analysis is the theory of #ground_rent. Land is not produced. Its price is a claim on future income, capitalised. Because location cannot be moved or copied, an owner of a well placed site can charge a payment simply for permitting access. That payment is rent, and it is a deduction from the incomes of others rather than a reward for producing anything. The rent gap idea, developed in the Marxist urban tradition, states that inner areas lose value as buildings age and capital is withdrawn, so that the rent actually collected falls well below the rent that could be collected if the land were used at its highest and best use. The gap between actual and potential rent is the fuel of reinvestment. When the gap is wide enough, and when institutions exist to bridge it, capital returns. This is the #rent_gap, and it explains why reinvestment concentrates in devalued areas rather than in already prosperous ones. The link to enclosure becomes obvious once we ask a simple question: who is standing on the land while the gap is being closed? The answer is: the people whose low rents constitute the actual rent. Closing the gap therefore requires either that they pay much more or that they leave. This is the analytical heart of the paper. Displacement is not a by product of closing the rent gap. Displacement is how the rent gap is closed. The tenant paying a low, protected or informal rent is, from the standpoint of capital, an obstacle occupying a site whose income potential is unrealised. Removing the tenant is not incidental to the investment. It is the investment. 2.4 Rentier capital and the asset economy The final theoretical element concerns where the money goes. Recent political economy has emphasised that a growing share of income in rich economies is captured through the ownership of scarce assets rather than through production. Christophers (2020) describes this as #rentier_capitalism, a system organised around the control of assets from which income can be extracted with little competition: land, buildings, spectrum, intellectual property, platforms, infrastructure. Adkins, Cooper and Konings (2020) argue that in an asset economy, position in the housing market rather than position in the labour market increasingly determines life chances. A person's income from work may matter less to their long term prospects than whether they owned property before prices rose. If this is right, then the city becomes the central arena of class formation. Wages are set nationally and by sector. Housing costs are set locally and by asset markets. A household with the same job in two neighbourhoods can be comfortable in one and destitute in the other. Under such conditions, the enclosure of urban space is not a marginal issue of amenity. It is the main current mechanism by which wealth is transferred upward. That is the claim this paper tests against the empirical literature. 2.5 A note on the word violence Describing gentrification as violent invites the objection that no one is being shot. The objection misses the point. Marx's use of blood and fire referred to hangings, whippings, vagrancy laws and the burning of cottages, but the analytical weight of his argument does not depend on spectacular cruelty. It depends on the fact that the separation of people from their means of life is enforced. Enforcement is the criterion. A bailiff who executes a possession order does not need to be cruel. The order is backed by the state's monopoly of force, and if the household does not go, force will be applied. Scholars working on eviction and estate demolition describe this as un-homing: the slow stripping away of the security, memory and social relations attached to a dwelling, ending in a physical removal that is often experienced as a bereavement (Elliott-Cooper, Hubbard and Lees, 2020). The category of #symbolic_violence captures the accompanying process by which residents are told that their neighbourhood is failing, that their community is a problem, and that their removal is for their own good. 3. Literature Review 3.1 Gentrification research and its long argument Gentrification research has been divided for fifty years between explanations that emphasise consumption and explanations that emphasise production. Consumption based accounts stress the preferences of a new middle class: a desire for old buildings, urban culture, short commutes and diverse street life. Production based accounts stress the movement of capital: disinvestment, devaluation, and the return of investment once the gap between actual and potential rent becomes large enough. The second tradition is the one on which this paper builds, but it does so with a qualification. The two explanations are not symmetrical. Preferences can only be acted on if the buildings are cheap enough to buy, and buildings are only cheap enough to buy because they were previously devalued. Taste selects among opportunities that capital has created. The recent literature has moved beyond this debate in three directions. First, it has become global. The processes once described in London and New York are now traced in Santiago, Seoul, Shanghai, Istanbul, Lagos, Manila and Mumbai, with the important finding that in many of these cities the state, not the individual middle class buyer, is the leading actor, and that the population displaced is often occupying land informally rather than renting formally (Lees, Slater and Wyly, 2023). This is the argument for #planetary_gentrification, which insists that gentrification is a global urban strategy rather than an Anglo American curiosity. Second, it has become explicit about violence. Elliott-Cooper, Hubbard and Lees (2020) argue that the standard vocabulary of displacement is too weak, because it suggests movement rather than loss. They propose un-homing to describe the destruction of dwelling as a relation, not merely as a location. This work connects gentrification research directly to the enclosure literature, since the enclosures also destroyed a relation to land and not simply a place of residence. Third, it has become attentive to the state. Watt (2021) documents the demolition and rebuilding of council estates in London and shows that the promise of like for like rehousing is regularly broken, that decanting scatters communities across the city and beyond, and that the new developments contain fewer homes at social rent than the estates they replaced. Slater (2021) situates such projects within a broader analysis of how ignorance is manufactured about urban inequality, so that structural causes are recoded as the failings of the poor. Lees and Warwick (2022) trace how design theories that treat poor neighbourhoods as dangerous have travelled between countries and been used to justify demolition. Together this work establishes that #state_led_gentrification is not an anomaly but a core mode. 3.2 The financialisation of housing The second body of literature concerns the transformation of housing into a financial asset. The key insight is that a dwelling can be understood in two ways. As a home, it is a #use_value: shelter, security, a place to raise children. As an asset, it is a stream of future payments that can be discounted, securitised, packaged and traded. The financialisation literature tracks the growing dominance of the second understanding over the first. Aalbers and colleagues have shown how private landlordism was reborn out of financialised homeownership, with buy to let mortgages allowing small investors to acquire rental property using leverage (Aalbers, Hochstenbach, Bosma and Fernandez, 2021). More consequential is the entry of large institutions. Taylor and Aalbers (2024) analyse the fifteen largest listed residential real estate investment trusts and similar funds in Germany and the United States and describe what they call a feeding machine: states supply the properties through privatisation, private equity supplies the capital, and exchange traded funds supply the constant flow of investment that pushes these landlords to keep growing. The result is a landlord that must expand, because its share price depends on growth, and that must raise rents, because its yield is measured quarterly. In the United States, the post crisis #foreclosure wave allowed institutions to buy family houses in bulk. Government analysis confirms the basic history: large institutional investors emerged in the wake of the 2007 to 2009 financial crisis, buying foreclosed homes at auction and converting them into rental housing, with concentrations in southern metropolitan areas (United States Government Accountability Office, 2024). Research on the consequences finds that concentrated institutional investment in #single_family_rentals reduced homeownership in affected neighbourhoods, with the sharpest effects on Black households in the Atlanta metropolitan region (An, 2024). Fields and Vergerio (2022) show how market power is built through geographic concentration and through fee income, and how technology allows remote landlords to manage thousands of scattered properties without local presence. Eviction research is the empirical hinge. Seymour and Akers (2021) show that in Detroit speculative landlords built a business model in which eviction filings were routine, not exceptional. Gomory (2022) demonstrates that eviction practices vary systematically by the institutional character of the landlord rather than only by tenant behaviour. Studies of the effects of investor purchases on rents and prices find upward pressure, though they debate magnitudes and mechanisms (Garriga, Gete and Tsouderou, 2023; Ganduri, Xiao and Xiao, 2023). Christophers (2023) provides the widest frame, arguing that asset managers now own the physical foundations of daily life, from pipes to homes, and that their time horizons and fee structures push them to extract value quickly and to underinvest in maintenance. 3.3 Public space, greening and the urban commons The third literature concerns shared space. Its central finding is that improvement of the environment, when carried out under conditions of private land ownership, tends to be capitalised into rent, and therefore tends to expel the people it was supposed to benefit. Anguelovski, Connolly and colleagues (2022) examined green interventions in a large sample of European and North American cities and found that in most of them greening was associated with gentrification, which they call #green_gentrification. Their later synthesis clarifies that the relationship is not automatic but depends on how greening interacts with real estate strategy and housing protections (Anguelovski and Connolly, 2024). Garcia-Lamarca and colleagues (2022) show that developers themselves market green amenities as a selling point, a practice they call urban #green_grabbing, in which environmental goods are enclosed as private assets. Earlier work in the same programme showed how green boosterism raises city affordability problems even while promoting sustainability credentials (Garcia-Lamarca and others, 2021). Alongside greening, tourism has become a powerful enclosure mechanism. Cocola-Gant and Gago (2021) trace how short term rental platforms in Lisbon combined with buy to let investment to convert residential buildings into tourist accommodation, driving displacement in central neighbourhoods. The mechanism is instructive: no law is needed to expel the tenant, only a change in the relative profitability of two uses of the same building. Platform intermediaries make this switch cheap and fast, and Sadowski (2020) describes the resulting model as an internet of landlords, in which digital infrastructure allows rent to be extracted from more and more of everyday life. Finally, the historical literature reminds us that planning as a discipline has a long relationship with the destruction of collective uses of space. Sevilla-Buitrago (2022) argues that modern planning emerged in part as a technique for dismantling the commons and for producing populations that could be counted, separated and governed. This is a strong claim, and it is not necessary to accept all of it to see its relevance. The instruments used to close a common field, to clear a slum and to redevelop an estate belong to the same family. 3.4 The gap in the literature Each of these bodies of work is strong. What is missing is an account that holds them together. Gentrification research often stops at the neighbourhood; finance research often stops at the balance sheet; public space research often stops at the plaza. As a result, policy responses are fragmented: an affordability quota here, a tenancy reform there, a public space charter somewhere else. If the three processes are moments of a single enclosure, then the fragmentation of the response is itself part of the problem. This paper attempts the synthesis. 4. Methodology 4.1 Research design This is a conceptual and interpretive study built on a structured synthesis of secondary literature. It does not generate new primary data. Its purpose is theoretical integration and reinterpretation: taking evidence that has already been produced by empirical researchers and reading it through a single analytical framework in order to test whether that framework holds. This choice of design follows from the research questions. The questions are not about the magnitude of an effect in a particular city. They are about whether a set of processes share a common structure. That kind of question is answered by conceptual work disciplined by evidence, not by regression. 4.2 Sources and selection The evidence base consists of peer reviewed articles and scholarly books, supplemented by a small number of official reports where they establish basic facts about ownership. Sources were selected according to four criteria. First, recency. Priority was given to work published within the last five years, because the ownership structure of urban housing changed rapidly after 2015 and older studies describe a different landlord population. Foundational theoretical texts are the exception, since the argument requires engagement with Marx and with Harvey. Second, empirical grounding. Preference was given to studies that report measured outcomes such as eviction filings, ownership shares, rent changes, tenure change or displacement counts, rather than to purely programmatic statements. Third, geographic spread. The synthesis draws on North American, European, Asian, African and Latin American cases, in order to avoid the well known bias of gentrification research toward a small number of Anglophone cities. Fourth, disconfirmation. Work that questions or qualifies the dispossession thesis was deliberately included, since a framework that only reads confirming evidence proves nothing. 4.3 Analytical procedure Each case discussed in Section 5 was examined against the three tests set out in Section 2.2: separation, extra economic force, and conversion into a revenue stream. Cases meeting all three tests were coded as enclosure. Cases meeting only some were retained and discussed as limits of the concept, because these limits are analytically informative. A second coding pass identified the specific instrument through which force operated: eviction, demolition, compulsory purchase, criminalisation, reclassification, design, licensing, or price. This produced the typology presented in Section 5.4. 4.4 Limitations of the design Three limitations should be stated at the outset. First, secondary synthesis inherits the biases of the underlying studies, including the over representation of certain cities. Second, the framework is explicitly theoretical, and a reader who rejects the labour theory of value and the associated theory of rent will reject part of the argument regardless of the evidence. This is acknowledged rather than concealed. Third, the paper cannot settle causal questions about the net effect of investment on neighbourhood outcomes, and it does not attempt to. Its claim is about mechanism, not about aggregate welfare. 5. Analysis 5.1 Gentrification as enclosure Consider a low income inner neighbourhood in almost any large city. Its buildings are old. Its rents are low, either because tenancy law protects them, because the buildings are in poor condition, or because the area carries a stigma. Some residents own outright, having bought decades ago. Some rent informally. Some occupy buildings without documents. Around them are institutions built through use rather than through purchase: a market, a mosque or church, a football pitch, a stretch of pavement where people sell food, a shortcut through a yard. From the point of view of an investor, this neighbourhood contains a large unrealised sum. Its land could support a much higher rent than it currently produces. The distance between those two figures is the #rent_gap, and it is the target. The gap can be closed in several ways, and the choice among them is a political choice, not a technical one. Market led closure. Individual buyers purchase and renovate. Rents rise as vacancies are re let at higher prices. Tenants without protection are displaced through the ordinary operation of tenancy law. Owner occupiers may sell at a profit, so this route produces some local winners. It is slow and it is uneven. State led closure. The state declares the area blighted, deprived, unsafe or underperforming. It compulsorily purchases, demolishes, and transfers the land to a developer, often at below market value, on condition that a proportion of the new units be affordable. This route is fast, and it is the dominant mode in much of Europe, East Asia and the global South. Watt (2021) shows how it operates in London through #estate_regeneration, where residents are decanted with promises of return that frequently fail. The number of homes at social rent falls, while the total number of units rises, and the arithmetic of the scheme depends on that substitution. Speculative closure. Land is acquired in advance of any development, held, and traded on the expectation of rezoning. Nothing is built for years. The mere expectation of future use raises current price and puts pressure on sitting tenants. This is characteristic of #speculative_urbanism in fast urbanising regions. Symbolic closure. Before any of the above, the area must be represented as a problem. Media, policy documents and even academic work portray it as failing, criminal, backward or empty. Slater (2021) shows how this production of ignorance prepares the ground: once a place is defined as a problem, its clearance becomes a solution rather than a seizure. Apply the three tests. Separation: residents lose access to the space. Force: the loss is enforced through eviction, demolition or compulsory purchase, and even in market led cases through possession orders backed by the courts. Conversion: the site afterwards yields a higher stream of rent to an owner. All three tests are met. Gentrification, in its state led and speculative forms above all, is enclosure. Two clarifications are needed. First, not every neighbourhood change is enclosure. Where an area's population changes gradually through voluntary moves, without force and without loss of access, the vocabulary does not apply. Second, enclosure does not require malice. The planner who approves the scheme may believe sincerely that the estate is unsafe and that new housing is better. The point of the analysis is structural, not moral. The system rewards a certain outcome regardless of the intentions of the people who produce it. 5.2 Corporate housing acquisition and the manufacture of the tenant The most important change in urban housing in the last fifteen years is not architectural. It is a change in who owns the buildings. Before the financial crisis, rental housing in most countries was owned by individuals and small firms. Landlords held a handful of units. They lived nearby. Their capacity to raise rents was limited by local knowledge, by inertia, and by the simple fact that finding a new tenant was troublesome. After the crisis, this changed. Millions of homes fell into foreclosure. They were sold in bulk, at auction, cheaply, and often to buyers with cash at a moment when ordinary buyers could not obtain mortgages. Public analysis confirms this pattern: institutional investors bought foreclosed homes at scale and converted them into rentals, concentrated in specific metropolitan areas of the American South (United States Government Accountability Office, 2024). This was #accumulation_by_dispossession in its clearest modern form. The dispossession happened first, through foreclosure, which was itself the consequence of predatory lending that had targeted the same neighbourhoods. The accumulation happened second, when the dispossessed homes were purchased by firms that then rented them back, sometimes to the very families who had lost them. A household moved, in a few years, from owner to tenant, from equity holder to fee payer, from a position where housing costs fell over time to a position where they rise every year. The wealth did not vanish. It moved. Three features distinguish the corporate landlord from the small one, and each of them intensifies enclosure. Scale and market power. When one owner holds a large share of the rental stock in a submarket, the tenant's option of moving elsewhere weakens. Fields and Vergerio (2022) document deliberate geographic concentration, which produces exactly this effect, and show that revenue is built not only from rent but from fees: for maintenance, for pets, for late payment, for utilities, for services the tenant did not choose. Distance and automation. The corporate landlord does not know the tenant and does not need to. Property technology allows remote assessment, remote pricing and remote enforcement. Sadowski (2020) describes how digital platforms transform ordinary property into an instrument for continuous rent extraction. Where rent setting is delegated to software that reads the market and recommends increases, the human hesitation that once moderated rent rises disappears. #algorithmic_rent_setting has become the subject of litigation precisely because it can coordinate increases across nominally competing landlords. The imperative to grow. A listed landlord is valued by capital markets on the basis of growth in income. It cannot decide to be content. Taylor and Aalbers (2024) describe the mechanism through which capital is fed into residential real estate investment trusts by exchange traded funds tracking indices, generating a structural pressure toward continual expansion. Christophers (2023) adds the crucial point about time: an asset manager operating a closed end fund must return capital to investors within a defined period, which encourages extraction now and discourages maintenance whose benefits fall due later. The consequences are measurable. Concentrated institutional purchase of family homes reduces homeownership in affected neighbourhoods, and the burden falls disproportionately on Black households (An, 2024). Investor purchases exert upward pressure on prices and rents (Garriga, Gete and Tsouderou, 2023). Eviction becomes a management tool: Seymour and Akers (2021) show a business model in Detroit in which filing for eviction was a routine step in cash flow management, and Gomory (2022) demonstrates that landlord type shapes eviction practice independently of tenant conduct. Apply the three tests again. Separation: the household loses the dwelling, first through foreclosure, later through eviction or unaffordable increase. Force: courts, bailiffs, possession orders. Conversion: the dwelling becomes a yielding asset in a portfolio. The tests are met. But there is a further point that distinguishes this case from the classical enclosure and makes it more insidious. In the historical enclosure, the peasant was expelled and the land was used for something else. In the corporate acquisition of housing, the tenant is often not expelled at all. The tenant is kept in place and converted into a source of income. The house remains a house. The family remains in the family home. What changes is that their continued residence is now the mechanism through which a distant fund earns a return. Enclosure here does not empty the space. It captures the person inside it. This is why the concept of #housing_as_asset is not a metaphor: the tenant is, in the language of the balance sheet, part of the asset, and the security of that tenancy is a variable to be optimised. 5.3 The privatisation of urban public space The third face of enclosure concerns the space between buildings. Its importance is easy to underestimate, because the loss is not usually recorded anywhere. Nobody is evicted from a square. Return to the logic of the commons. The English peasant did not simply live on common land. The peasant used it: for grazing, for fuel, for gleaning, for gathering. These uses were not luxuries. They were the margin that made a low cash income survivable. When they were extinguished, subsistence became fully monetised, and the same wage that had been enough became insufficient. Urban shared space performs the same function today. A public library is warmth and internet access. A public toilet is dignity and the ability to spend a day in the city without money. A drinking fountain is water. A park bench is rest. An unpoliced street corner is the possibility of informal trade. A cheap street market is food at prices that no supermarket matches. A shortcut across an estate is time. Collectively, these things constitute an urban commons, and they are the reason it is possible to be poor in a city and still live. The enclosure of this commons takes several forms. Direct privatisation. Squares, plazas and open spaces are transferred to private management, often as part of a development agreement. They remain visually open but become #privately_owned_public_spaces, governed by private rules, patrolled by private security, and subject to conditions of entry that need not be justified in law. Photography may be prohibited, protest is normally prohibited, and behaviour that is legal on a public street can be grounds for exclusion. Design based exclusion. Benches acquire armrests that prevent lying down. Ledges acquire studs. Doorways acquire spikes. Sprinklers run at night. This is #hostile_architecture, and its function is precise: to make certain uses of space impossible without making them illegal, thereby avoiding the need for any democratic decision at all. Commercial occupation. Streets become terraces. Squares become event venues. Access is not forbidden, but it becomes conditional on consumption. The person who is not buying is not excluded by rule, only by the absence of anywhere to be. Green enclosure. Parks, waterfronts and greenways are created or improved, environmental quality rises, and property values rise with them. Anguelovski and colleagues (2022) find that in most of the cities they studied greening was accompanied by gentrification. Developers actively use environmental amenity as a marketing asset, a practice described as urban #green_grabbing (Garcia-Lamarca and others, 2022). The cruelty of this pattern is that the residents who campaigned for a cleaner environment, often after decades of exposure to industry and traffic, are the ones who cannot afford to stay once it arrives. Tourist enclosure. Whole districts are reorganised around visitors. Housing becomes accommodation. Grocers become souvenir shops. Cocola-Gant and Gago (2021) trace this process in Lisbon and show the mechanism plainly: when short term letting to tourists is more profitable than long term letting to residents, buildings switch use, and no law is needed to accomplish the expulsion. #tourism_gentrification is enclosure by price, executed through platform infrastructure. Criminalisation. Finally, the direct route. Begging, sleeping, loitering, informal vending and gathering are made offences. The #policing_of_the_poor removes people from valuable space without requiring anyone to buy the space. This is the purest modern descendant of the vagrancy statutes that Marx described, in which the very people created by expropriation were punished for the condition that expropriation had produced. Apply the tests. Separation: poor residents lose the use of space they had used. Force: private security, byelaws, design, criminal law. Conversion: the surrounding land yields higher rent, and in the case of privatised plazas the space itself yields revenue. The tests are met. The #urban_commons is being enclosed, and the effect on the cost of urban survival is the same as the effect of the extinction of grazing rights three centuries ago. 5.4 The machinery: how force is applied without appearing as force The strength of the enclosure analogy depends on identifying the instruments. Marx could point to specific acts of parliament. What are the modern equivalents? The synthesis produces a typology of seven instruments, each of which appears in the cases above. One: legal expropriation. Compulsory purchase, eminent domain, land readjustment, and their equivalents. The state takes the land, pays compensation, and transfers it. Compensation is calculated at existing use value, which is precisely the value that the project is designed to increase, so the former owner is systematically excluded from the gain that the taking creates. Two: tenure reclassification. Informal settlements are declared illegal, so that their inhabitants have no claim; or protected tenancies are converted to unprotected ones through reform; or public housing is reclassified as an asset to be disposed of. A stroke of the pen removes a right that had existed in practice for decades. Three: viability. A development is assessed for its financial viability, and obligations to provide affordable housing are reduced on the ground that they would render the scheme unprofitable. The assessment treats the developer's expected return as a fixed input and the public benefit as the adjustable one. This is an administrative procedure, not a political decision, and it is therefore very hard to contest. Four: debt. Foreclosure, arrears, service charges levied on leaseholders after regeneration, and mortgage default transfer property from households to institutions without any coercive act that looks like coercion. The bailiff arrives at the end of a process that began with a signature. Five: design and byelaw. Exclusion accomplished by the physical form of space and by minor regulation, below the threshold at which anyone would think to object politically. Six: price. The most common instrument, and the one that appears most natural. A rent increase requires no order and no officer. But the tenant who cannot pay must go, and the removal is enforceable. Price is force at one remove. Seven: narrative. The prior representation of a place as failing, dangerous, empty or backward, which makes all the other instruments appear reasonable. The seven instruments have a common property. Each one converts a political question, namely who has the right to remain in the city, into a technical or private one. This is the mechanism by which contemporary enclosure achieves what parliamentary enclosure achieved with statutes: it makes dispossession procedurally correct. 5.5 Where the value goes The final analytical step is to follow the money. If enclosure produces a gain, who receives it? The gain from closing a rent gap takes the form of increased land value and increased rental income. It accrues to whoever owns the land at the moment the gap closes. Not to the construction workers, whose wages are set by the labour market. Not to the tenants, who pay more. Not usually to the municipality, which in most systems captures only a fraction through taxation and planning obligations, and which frequently sells its own land cheaply to make the scheme happen at all. The gain accrues to owners: to developers, to landowners, to investors and, increasingly, to the funds that stand behind them. This is what makes the process rentier rather than productive. New value is created when a house is built, in the form of the labour and materials embodied in it. But the far larger sum captured through the appreciation of land is not created by anyone. It is a claim on the future incomes of the people who will have to live there. Christophers (2020) is precise on this point: rentier income depends on control of a scarce asset rather than on the production of anything, and its growth is therefore a redistribution rather than an addition. The urban consequence is a redistribution from wage earners to asset holders, executed through the housing system, and it operates continuously rather than in a single act of theft. Adkins, Cooper and Konings (2020) argue that this reorganises class itself: the decisive division is no longer only between those who own capital and those who sell labour, but between those who own housing assets and those who must rent them. A generation locked out of ownership pays a rising share of its wages to a generation and a class that owns, and the gap compounds. #uneven_development, in the twenty first century, runs as much between tenures as between regions. 6. Discussion 6.1 Objections to the argument An argument of this kind should be tested against the strongest available objections, and there are several. Objection one: gentrification improves neighbourhoods, and improvement is good. Reinvestment brings repaired buildings, safer streets, better shops, better schools, cleaner air. Some longstanding residents benefit. Owner occupiers gain equity. Empirical work on the effects of neighbourhood change is genuinely mixed, and some studies find that many original residents stay and experience improved conditions. The response is not to deny the improvements but to ask who pays for them and who receives them. The analysis in this paper does not claim that nothing improves. It claims that the improvement is captured as rent, and that the capture requires the removal of those who cannot pay. If a park makes an area pleasant and the pleasantness is charged for through rent, then the poor residents have not received a park. They have received an eviction notice with a longer lead time. The test is simple and it is empirical: does the improvement reach the people who lived there before, or does it reach whoever can outbid them? Where strong tenant protection, #social_housing and land #value_capture exist, improvement can reach existing residents. Where they do not, it cannot. The variable is not investment. The variable is who owns the ground. Objection two: the concept of primitive accumulation is being stretched beyond usefulness. If every transfer of value is called dispossession, the term explains nothing. This objection is serious and it is the reason for the three tests set out in Section 2.2. A rent rise in a competitive market with a well protected tenant who chooses to move is not enclosure. A landlord who sells to another landlord is not enclosure. What distinguishes enclosure is the combination of separation, extra economic force and conversion into a revenue stream. Applied honestly, this excludes a large amount of ordinary market activity, and it should. Objection three: displacement is hard to measure, and some studies find less of it than expected. This is true and important. Displaced households are, by definition, absent from the area being surveyed. They are hard to trace. Some quantitative work finds mobility rates in gentrifying areas that are not much higher than elsewhere. But three qualifications matter. Households can be displaced in place, remaining in the dwelling while losing the neighbourhood that made it liveable. Households can be prevented from arriving, a form of exclusionary displacement that leaves no trace in any survey of movers. And households can be displaced across a generation, as children who grew up in an area find that they cannot afford to live in it. The measured mobility rate captures none of these. Objection four: the state is not a simple agent of capital. True. States build social housing, impose rent regulation, and sometimes block developments. The relationship is contradictory, and the analysis presented here would be crude if it treated the state as a mere instrument. What can be said is narrower and stronger: where public authorities depend on land sales, property taxes and development levies for their revenue, they acquire a direct financial interest in rising land values, and that interest shapes what they can imagine doing. #austerity_urbanism creates a municipality that must monetise its own ground to survive. Objection five: the global South does not fit the model. In many cities the displaced are not formal tenants but informal occupants; the actor is not a middle class buyer but the state; and the land in question was never a commons in the European sense. This is correct, and it is why planetary approaches to gentrification insist on the specificity of context (Lees, Slater and Wyly, 2023; Ibrahim and others, 2024). But the specificity strengthens rather than weakens the enclosure reading. Where residents hold land by custom, occupation or long tolerance rather than by title, they are in a position structurally closer to the commoner of the sixteenth century than any Western tenant is, and their expulsion by bulldozer for a business district is closer to the original enclosure than anything happening in London. 6.2 What is genuinely new Fidelity to the historical analogy requires also identifying what has changed. Four differences deserve emphasis. First, the enclosure is now often invisible, because it is executed through price and through ownership change rather than through fences. A fund can acquire an entire street without a single resident noticing until the letters arrive. Second, the dispossessed are frequently retained rather than expelled. As argued in Section 5.2, the corporate landlord's aim is not an empty house. It is an occupied house with a rising rent. Enclosure that keeps the person inside the enclosure is a form the sixteenth century did not know. Third, the beneficiary is distant and diffuse. The landlord is a fund; the fund is owned by other funds; the other funds hold the pensions of ordinary workers. There is no squire. This is not an argument against the analysis, but it makes resistance harder, because there is no gate to break. Fourth, the enclosure now presents itself as environmental virtue. The green waterfront, the tree lined street and the flood resilient park are genuinely good things, and they are being used as instruments of rent extraction. This is a novelty and a trap, because it forces communities into the impossible position of opposing improvements they need. 6.3 Theoretical implications If the argument holds, three theoretical adjustments follow. The first concerns the periodisation of capital. Primitive accumulation should be treated not as a stage but as a function, one that is reactivated whenever the returns from production weaken and asset ownership offers better returns. On this reading, the last four decades of urban policy are not a deviation from capitalism but a return to its founding technique. The second concerns the theory of the working class. In classical accounts, workers are exploited at the point of production, through the extraction of surplus value from labour. In the asset economy, a second extraction occurs at the point of residence, through rent. A worker may be exploited twice: once by the employer and once by the landlord who is, increasingly, an arm of the same financial system. Wage struggle alone cannot address the second extraction, because rent absorbs wage gains wherever housing supply is controlled by rentiers. This is a direct argument for treating #tenant_organizing as a form of class organisation rather than as a consumer complaint. The third concerns planning theory. If planning's instruments regularly function to enclose, then planning cannot be understood as a neutral technique for balancing interests. Sevilla-Buitrago (2022) makes the historical case that the discipline emerged partly to dismantle collective uses of land. Contemporary planners are not obliged to accept this inheritance, but they cannot pretend it does not exist. The choice between viability led development and #decommodification is a political choice made in a technical register. 7. Implications for Policy and Planning If displacement is a mechanism of accumulation rather than an accident, then policy must target the mechanism. The following measures follow from the analysis rather than from a general preference for intervention. They are grouped by what they act upon. 7.1 Acting on ownership Public and social housing at scale. Housing removed from the market cannot be enclosed. This is the simplest implication of the whole analysis and it is the one most often avoided. A dwelling owned by a public body, a cooperative or a non profit trust, and let at a rent tied to income rather than to land value, is a dwelling from which nobody can be expelled in order to capture a rent gap, because there is no rent gap to capture. Community land trusts. Separating ownership of the land from ownership of the building removes the land from speculation permanently. The building can be sold; the land cannot. The #community_land_trust is the direct institutional negation of enclosure, and it works precisely because it restores a form of collective tenure. Limits on institutional acquisition. Several jurisdictions have moved to restrict bulk purchases of family homes by entities holding large portfolios, and to prohibit the packaging of multiple homes into single transactions designed for institutional buyers. Given the evidence that concentrated institutional ownership reduces homeownership and raises rents (An, 2024; Garriga, Gete and Tsouderou, 2023), such restrictions target a documented harm. Right of first refusal. Where a rental building is sold, tenants or a public body are given the first opportunity to purchase at the market price. This converts a moment of maximum vulnerability into a moment of possible acquisition. 7.2 Acting on rent Rent regulation. Regulation is contested among economists, and honest treatment requires acknowledging the standard criticism that binding controls can reduce the supply and quality of rental housing. The counter argument, in the terms of this paper, is that unregulated rent in a market of scarce land is not a price signal that calls forth supply but a transfer to landowners, since the supply of urban land is fixed by definition. The empirical question is contested; the theoretical question about land is not. Regulation of algorithmic pricing. Where landlords set rents using shared software that draws on the same market data, the result approximates coordination even without a meeting. This deserves scrutiny under competition law rather than under housing law. Just cause eviction and long tenancies. If eviction is the instrument through which the tenant is separated from the dwelling, then restricting the grounds for eviction directly disables the instrument. Given the evidence that eviction functions as a routine management tool for some corporate landlords (Seymour and Akers, 2021; Gomory, 2022), this is a targeted response. 7.3 Acting on land value Value capture. When public investment raises land values, the increase is created by the public and captured by owners. Instruments that recover a substantial share of that increase, whether through land value taxation, betterment levies or public land leasing, break the link between public improvement and private windfall. This is the only way to make greening, transit and public space investment safe for the people who currently live near them. Public land banking. A public authority that holds land can develop it without paying speculative prices to itself. Authorities that sold their land during earlier decades of privatisation now buy it back at prices they created. 7.4 Acting on the commons Legal protection of public space. Privately owned public spaces should be subject to the same rules of access, assembly and behaviour as public streets, and any transfer of public space to private management should require the retention of public rights in law. Investment in the infrastructure of low income survival. Public toilets, drinking water, libraries, benches without armrests, street markets, unpoliced places to sit. These are cheap, they are unglamorous, and they are the material basis on which it is possible to be poor in a city without being destitute. Their removal is a rent increase by other means. Anti displacement conditions attached to greening. Environmental improvement should be planned together with protection of the residents who will benefit from it, through preservation of affordable stock, tenant protections and land trusts established before the improvement is delivered, not after (Anguelovski and Connolly, 2024). 7.5 Acting on power None of the above happens without organised pressure, because each measure reduces a stream of income currently flowing to well organised interests. The literature on housing movements documents tenant unions, eviction blockades, rent strikes and campaigns for municipal acquisition as the political condition of policy change (Rosenthal and Vilchis, 2024). The analytical point is that these movements are not lobbying for a concession within the system. They are contesting an enclosure, and their historical predecessors are the commoners who pulled down fences. 8. Limitations and Future Research This study has clear limits. It is a synthesis and not an original empirical investigation, so it cannot adjudicate the quantitative disputes it describes, particularly the contested magnitude of investor effects on rents and the contested measurement of displacement. The framework is theoretically committed, and readers who reject rent theory will reject part of the argument on prior grounds. The evidence base, despite efforts to widen it, remains weighted toward North America and Western Europe, and the specificities of Gulf, South Asian, African and Latin American urbanisation deserve their own treatment rather than assimilation to a model built elsewhere. Four lines of future work follow. First, comparative measurement of ownership concentration at the neighbourhood scale, which remains poor in most countries because ownership registries are fragmented or closed. Second, longitudinal tracing of displaced households, which is the single largest evidential gap in the field. Third, study of the interaction between climate adaptation spending and land values, since the next decade will see very large public investment in flood defence, cooling and green infrastructure, all of which will be capitalised into rent unless value capture is in place. Fourth, evaluation of counter enclosure institutions, especially community land trusts and public acquisition programmes, which are frequently advocated and rarely assessed at scale. 9. Conclusion Marx ended the first volume of Capital with the history of expropriation because he wanted to destroy a story. The story said that capital was the reward of virtue and that the propertyless had only themselves to blame. Against it he set the fence, the statute and the burnt cottage. The urban story told today is a descendant of the one Marx attacked. It says that neighbourhoods change because people's tastes change; that rents rise because demand rises; that those who cannot afford to stay have simply been outbid; that investment is good and that displacement, regrettable as it is, is the price of improvement. This paper has argued that the story hides the mechanism. Contemporary urban development does not merely coincide with the removal of working class residents from valuable land. In a wide class of cases it depends on that removal, because the profit sought is the difference between what the land yields now, with those residents on it, and what it could yield without them. The three processes examined here are one process. #urban_gentrification separates residents from central land. #corporate_landlords separate households from ownership and convert their tenancy into an income stream. The #public_space_privatization of streets, squares and parks separates the poor from the shared resources that make low incomes survivable. Each meets the tests of separation, of extra economic force, and of conversion into rent. Each is defended by an appeal to improvement. Each transfers value from those who work to those who own. The instruments are compulsory purchase and viability assessment rather than parliamentary enclosure acts, and the beneficiary is a fund rather than a squire, but the operation is the one Marx described. The practical implication is uncomfortable for planning as it is currently practised. A policy that seeks to attract investment while mitigating displacement is attempting to keep the effect while removing the cause. If the analysis here is right, the only measures that reach the mechanism are those that take land and housing out of the circuit of #rentier_capital: public and cooperative ownership, land trusts, value capture, secure tenancies, and the defence of the commons as a material condition of urban life rather than as an amenity. The enclosure of the commons was not a natural process. It was fought, in the fields and in the courts, and it was fought for two centuries. Its urban successor is being fought now, in eviction courts, in tenant unions, in campaigns against demolition, and in the ordinary refusal of people to leave the places where they live. Whether the twenty first century city becomes a machine for the extraction of #ground_rent or a place that people can afford to inhabit is not a technical question about supply. It is a question about who owns the ground, and that has always been a question of power. Hashtags #urban_geography_research #planning_theory #marxist_urban_studies #gentrification_and_displacement #housing_financialisation #enclosure_of_urban_space #rentier_economy #critical_urban_theory #right_to_the_city #housing_justice_now #land_and_power #cities_and_capital #tenant_power #anti_displacement #urban_political_economy References Aalbers, M. B., Hochstenbach, C., Bosma, J. and Fernandez, R. (2021) The death and life of private landlordism: how financialized homeownership gave birth to the buy to let market. Housing, Theory and Society, 38(5), 541 to 563. Adkins, L., Cooper, M. and Konings, M. (2020) The Asset Economy: Property Ownership and the New Logic of Inequality. Cambridge: Polity Press. An, B. Y. (2024) The influence of institutional single family rental investors on homeownership: who gets targeted and pushed out of the local market? Journal of Planning Education and Research, 44(4), 2054 to 2071. Anguelovski, I. and Connolly, J. J. T. (eds.) (2022) The Green City and Social Injustice: 21 Tales from North America and Europe. Abingdon: Routledge. Anguelovski, I. and Connolly, J. J. T. (2024) Segregating by greening: what do we mean by green gentrification? Journal of Planning Literature, 39(1), 21 to 35. Anguelovski, I., Connolly, J. J. T., Cole, H., Garcia-Lamarca, M., Triguero-Mas, M., Baro, F., Martin, N., Conesa, D., Shokry, G., Perez del Pulgar, C., Arguelles Ramos, L., Matheney, A., Gallez, E., Oscilowicz, E., Lopez Manez, J., Sarzo, B., Beltran, M. A. and Martinez Minaya, J. (2022) Green gentrification in European and North American cities. Nature Communications, 13, 3816. Blakeley, G. (2024) Vulture Capitalism: Corporate Crimes, Backdoor Bailouts and the Death of Freedom. London: Bloomsbury. Christophers, B. (2020) Rentier Capitalism: Who Owns the Economy, and Who Pays for It? London: Verso. Christophers, B. (2023) Our Lives in Their Portfolios: Why Asset Managers Own the World. London: Verso. Cocola-Gant, A. and Gago, A. (2021) Airbnb, buy to let investment and tourism driven displacement: a case study in Lisbon. Environment and Planning A: Economy and Space, 53(7), 1671 to 1688. Colburn, G. and Aldern, C. P. (2022) Homelessness Is a Housing Problem: How Structural Factors Explain U.S. Patterns. Oakland: University of California Press. Elliott-Cooper, A., Hubbard, P. and Lees, L. (2020) Moving beyond Marcuse: gentrification, displacement and the violence of un-homing. Progress in Human Geography, 44(3), 492 to 509. Fields, D. and Vergerio, M. (2022) Corporate Landlords and Market Power: What Does the Single Family Rental Boom Mean for Our Housing Future? Berkeley: University of California, Berkeley, Institute of Governmental Studies. Ganduri, R., Xiao, S. C. and Xiao, S. W. (2023) Tracing the source of liquidity for distressed housing markets. Real Estate Economics, 51(2), 408 to 440. Garcia-Lamarca, M., Anguelovski, I., Cole, H., Connolly, J. J. T., Arguelles, L., Baro, F., Perez del Pulgar, C. and Shokry, G. (2021) Urban green boosterism and city affordability: for whom is the branded green city? Urban Studies, 58(1), 90 to 112. Garcia-Lamarca, M., Anguelovski, I., Cole, H. V., Connolly, J. J. T., Perez del Pulgar, C., Shokry, G. and Triguero-Mas, M. (2022) Urban green grabbing: residential real estate developers discourse and practice in gentrifying Global North neighborhoods. Geoforum, 128, 1 to 10. Garriga, C., Gete, P. and Tsouderou, A. (2023) The economic effects of real estate investors. Real Estate Economics, 51(3), 655 to 685. Gomory, H. (2022) The social and institutional contexts underlying landlords eviction practices. Social Forces, 100(4), 1774 to 1805. Harvey, D. (2003) The New Imperialism. Oxford: Oxford University Press. Ibrahim, A.-S., Abubakari, M., Cobbinah, P. B. and Kuuire, V. Z. (2024) Accumulation by dispossession and the truism of urban regeneration. Journal of Planning Literature, 39(4), 553 to 568. Lees, L. and Warwick, E. (2022) Defensible Space on the Move: Mobilisation in English Housing Policy and Practice. Chichester: Wiley Blackwell. Lees, L., Slater, T. and Wyly, E. (eds.) (2023) The Planetary Gentrification Reader. Abingdon: Routledge. Marx, K. (1976) Capital: A Critique of Political Economy, Volume I. Translated by B. Fowkes. London: Penguin Classics. Originally published 1867. Rosenthal, T. and Vilchis, L. (2024) Abolish Rent: How Tenants Can End the Housing Crisis. Chicago: Haymarket Books. Sadowski, J. (2020) The internet of landlords: digital platforms and new mechanisms of rentier capitalism. Antipode, 52(2), 562 to 580. Sevilla-Buitrago, A. (2022) Against the Commons: A Radical History of Urban Planning. Minneapolis: University of Minnesota Press. Seymour, E. and Akers, J. (2021) Building the eviction economy: speculation, precarity, and eviction in Detroit. Urban Affairs Review, 57(2), 414 to 442. Slater, T. (2021) Shaking Up the City: Ignorance, Inequality, and the Urban Question. Oakland: University of California Press. Taylor, Z. J. and Aalbers, M. B. (2024) The restless urban landscape of housing financialization: geographies of residential real estate investment trust expansion in Germany and the United States. Urban Geography, 45(9), 1822 to 1842. United States Government Accountability Office (2024) Rental Housing: Information on Institutional Investment in Single Family Homes. Report GAO 24 106643. Washington DC: GAO. Watt, P. (2021) Estate Regeneration and Its Discontents: Public Housing, Place and Inequality in London. Bristol: Policy Press. Wijburg, G. (2021) The de financialization of housing: towards a research agenda. Housing Studies, 36(8), 1276 to 1293.

This article studies the gig economy as a system for producing and capturing #surplus_value. It brings together the classical Marxist distinction between absolute and relative surplus value with recent empirical research on #algorithmic_management in ride hailing, food delivery and online freelance markets. The central claim is simple. Platforms have not invented a new way of making money out of work. They have found a cheaper and more precise way of doing something old. Through location tracking, dynamic pricing, ranking systems and automated allocation of tasks, platform firms extend the effective working day and raise the intensity of labour at the same time, while shifting the cost of tools, insurance, sick pay, pensions and idle time onto the worker. The result is a labour process in which the boundary between paid and unpaid time becomes blurred, and in which the rate of exploitation can be adjusted almost continuously and at the level of the individual worker. The article develops this argument in four steps. First, it restates the categories of absolute and relative surplus value and explains why they still apply to workers who are legally classified as #independent_contractor rather than employees. Second, it reviews the empirical literature on algorithmic control published mainly between 2020 and 2025. Third, it presents an integrative analysis of six mechanisms through which platforms increase surplus value: unpaid waiting time, intensification through routing and gamification, the externalisation of fixed costs, the capture of data as a second stream of value, personalised and opaque pay, and the management of worker resistance. Fourth, it discusses what this means for #labour_law, collective bargaining and future research. The article concludes that #platform_capitalism should be read less as a technological rupture and more as a deepening of the wage relation under conditions where the wage form itself has been legally dissolved. Keywords: algorithmic management; surplus value; gig economy; platform capitalism; labour process theory; precarious work; digital labour; employment classification 1. Introduction A driver in Dubai switches on an application at seven in the morning. She sits in her car for forty minutes before the first request arrives. She is not paid for those forty minutes. She accepts a trip that pays a fare calculated by a model she cannot see, based on demand conditions she cannot verify, and she knows that if she refuses too many trips her access to work may quietly shrink. At the end of the day she has been logged in for eleven hours, has driven for six, has been paid for five, and has covered the cost of fuel, insurance, maintenance and her own health cover out of what remains. Legally she is a business. Practically she is being managed, minute by minute, by software. This everyday scene is the empirical starting point of the present article. The scene is not confined to ride hailing. It repeats itself, with local variations, in food delivery, grocery courier work, domestic cleaning platforms, care platforms, microtask sites and online #freelance_marketplace. Across these settings a common architecture has emerged. Work is distributed by algorithm. Performance is measured continuously. Pay is variable and often personalised. Legal responsibility for the costs and risks of production is pushed downward onto the person doing the work. Firms describe themselves as neutral intermediaries that merely connect supply with demand, while retaining the power to set prices, allocate tasks, define quality and exclude workers from the market. Sociology and labour studies have documented these arrangements in considerable detail over the last decade. There is now a large body of qualitative and quantitative work on how couriers experience tracking, how drivers respond to ratings, how crowdworkers manage reputation, and how platform workers organise despite the absence of a shared workplace. What remains comparatively underdeveloped is a systematic account of why these arrangements exist in the form they take. Descriptive accounts of control tell us how workers are governed. They do not always tell us what the governing is for. This article argues that the purpose is the maximisation of #surplus_value, and that the classical distinction between absolute and relative surplus value remains the most economical way to explain the specific design choices platforms make. Absolute surplus value is obtained by lengthening the working day or by capturing time that was previously outside the wage relation. Relative surplus value is obtained by reducing the labour time necessary to reproduce the worker, typically by raising productivity and intensity. Platforms pursue both at once, and they do so through the same technical apparatus. The application that tracks a courier's position in order to assign the next order is simultaneously the instrument that keeps her waiting without pay, the instrument that pushes her to complete more drops per hour, the instrument that adjusts her pay downward when supply is abundant, and the instrument that produces the data used to refine all of the above. The argument proceeds through four claims. First, that the legal reclassification of workers as contractors is not incidental to the extraction of surplus value but central to it. By dissolving the employment contract, platforms dissolve the legal definition of the #working_day. Where there is no employer, there is no shift, no overtime, no rest break, no paid standby. Time that a factory or a supermarket would have to pay for becomes free. Second, that #dynamic_pricing functions as a mechanism of intensification. It is usually presented as an efficiency device that balances supply and demand. In labour terms it is a device that moves workers to where value can be realised fastest, that shortens the unproductive gaps between tasks, and that varies the price of labour power in real time according to the platform's needs rather than the worker's. Third, that the externalisation of vehicles, phones, data plans, fuel, insurance and social protection lowers the platform's capital outlay dramatically. The firm carries almost no #constant_capital in the traditional sense. Its fixed capital is code and data. The physical means of production are financed by the workers themselves, which raises the measured profitability of the platform without any change in the underlying productivity of the labour performed. Fourth, that data is not merely a by-product of platform work but a second stream of appropriation. Every completed task trains the systems that will later be used to allocate work more tightly, to price it more finely, and in some cases to reduce the skill required to perform it. The article is written for students and researchers in sociology, labour studies, political economy and management. It is deliberately written in plain language. Where technical terms from Marxist political economy are used, they are defined. The empirical material is drawn from published research rather than from new fieldwork, and the contribution is theoretical and integrative: it offers a framework that connects a fragmented empirical literature to a coherent account of value. The remainder of the article is organised as follows. Section 2 sets out the theoretical framework. Section 3 reviews the recent literature on algorithmic control and platform political economy. Section 4 explains the method. Section 5 presents the analysis in six parts. Section 6 discusses the implications for theory. Section 7 turns to policy. Section 8 notes limitations and directions for further work. Section 9 concludes. 2. Theoretical Framework 2.1 Surplus value in one paragraph In the Marxist account, the worker sells not labour but labour power, the capacity to work for a given period. The value of labour power is determined by what it costs to reproduce the worker at a socially accepted standard of living. If that cost is equal to four hours of average social labour, and the worker works for ten hours, then six hours are unpaid. Those six hours are the source of #surplus_value. The ratio between unpaid and paid labour time is the rate of exploitation. Nothing in this argument depends on the worker being badly treated, on the employer being greedy, or on the wage being unfair by some moral standard. It is a structural relation, and it holds even when wages rise, provided that productivity rises faster. Two paths lead to more surplus value. The first is to make the working day longer while the value of labour power stays the same. This is #absolute_surplus_value. The second is to reduce the portion of the day needed to reproduce the worker, usually by making the production of wage goods more productive, or by making the worker work harder within the same hours. This is #relative_surplus_value. Classical accounts treat intensification as a special case of the second path, because working more densely within an hour is functionally similar to lengthening the hour. The distinction matters because it points at different institutional battlegrounds. The struggle over absolute surplus value is a struggle over time: shift length, breaks, overtime, standby, on call arrangements. The struggle over relative surplus value is a struggle over the pace and organisation of work: how many units per hour, how tightly is movement scripted, how much discretion remains. The gig economy is interesting precisely because it reopens both struggles simultaneously, on ground where the legal instruments developed over a century of industrial conflict do not automatically apply. 2.2 The labour process and the indeterminacy of labour Labour process theory adds a crucial point. When a firm buys labour power it buys a potential, not a fixed quantity of output. The gap between what the worker could do and what the worker actually does is the indeterminacy of labour. Management exists, in the first instance, to close that gap. Every historical form of management, from direct supervision to the assembly line to bureaucratic career ladders, is a solution to the same problem: how to convert purchased potential into actual effort. This is why the study of control is not separate from the study of value. Control is the technical means by which surplus value is realised. When we study how a delivery application decides which courier receives which order, we are studying an instrument of value extraction, not merely an instrument of coordination. The classical typology distinguishes simple control by the direct authority of a supervisor, technical control embedded in machinery and the pace of the line, and bureaucratic control embedded in rules, hierarchies and internal labour markets. #Algorithmic_management does not fit neatly into any of these. It resembles technical control in that the pace is set by a machine. It resembles bureaucratic control in that rules are formalised and applied impersonally. It resembles simple control in that it is continuous, granular and personal in its effects. But it differs from all three in one decisive respect: it operates without the legal form of employment that historically accompanied them, and it does so at a distance, on workers who are formally free to refuse any given task. 2.3 Formal and real subsumption Two further terms are needed. Under #formal_subsumption, capital takes over a labour process that already exists without changing its technical content. The craftsman keeps his tools and his methods, but now works for a wage. Surplus value is raised mainly by extending hours. Under #real_subsumption, capital transforms the labour process itself. Machinery, division of labour and scientific management restructure the work so that the technical content of the job is shaped by the requirements of accumulation. Surplus value is raised mainly by raising productivity and intensity. The gig economy presents an unusual hybrid. Formally, the worker looks like an independent producer who owns the means of production, sets his own hours and bears his own risks. This is the appearance of a pre capitalist or petty commodity form. In substance, however, the labour process is deeply restructured. Routes are calculated, sequences are imposed, timing is monitored, quality is standardised, the price is set by the platform, and the customer relationship belongs to the platform, not the worker. The worker owns the car but does not own the market, does not own the customer, and does not own the information. The appearance of independence coexists with an intensity of control that exceeds many conventional workplaces. This is why several recent contributions describe platform work as combining the legal shell of formal independence with a substantive #real_subsumption of the labour process. 2.4 The piece wage returns The wage form used by most platforms is a piece wage, sometimes disguised. Payment is made per trip, per drop, per task, per accepted assignment, occasionally per hour but with acceptance and completion conditions attached. The classical analysis of the #piece_wage is directly relevant here, and it makes three observations that anticipate the gig economy with uncomfortable accuracy. First, the piece wage is the wage form that best suits capital because it makes quality and intensity self policing: the worker who slows down punishes herself. Second, it converts supervision into an internal function of the worker, since the worker now has an interest in extending her own hours. Third, it creates the appearance that the worker is paid for her product rather than for her time, which obscures the existence of unpaid labour time altogether. The platform piece wage adds a further twist that the classical analysis did not anticipate. In a nineteenth century workshop the rate per piece was known in advance and applied to everyone. On a contemporary platform the rate per task is calculated by a model, may differ between two workers doing identical work at the same moment, and is not disclosed. Recent legal scholarship names this #algorithmic_wage_discrimination and argues that it represents a qualitative change in the wage relation, because it allows the firm to identify and pay each worker close to the minimum amount that will induce that individual to accept the task. 2.5 The framework in summary Bringing these threads together, the analytical framework used in this article can be stated as five propositions. Platform firms extract surplus value from labour performed by workers who are legally classified as independent, and the classification is a condition of the extraction rather than a separate legal quirk. The extension of unpaid, available, logged in time is a strategy of absolute surplus value, made possible by the absence of an employment contract that would define and price the working day. Algorithmic allocation, routing and gamified incentives are strategies of relative surplus value, because they raise the density of effort within any given hour. The externalisation of fixed costs and social protection raises the profit rate without raising productivity, and functions as a transfer of the costs of #social_reproduction from capital to labour and to the state. The capture and processing of behavioural data constitutes a second circuit of appropriation, which feeds back into propositions two and three by improving the precision of both. These propositions are tested against the empirical literature in Section 5. 3. Literature Review 3.1 The rise of algorithmic management as a research field The study of software driven work coordination has moved, in roughly a decade, from a niche interest to a central topic in the sociology of work. Early contributions described how ride hailing firms used ratings, incentives and information asymmetries to steer drivers without formally instructing them. A widely cited synthesis published in 2020 organised the field around the observation that algorithms direct, evaluate and discipline workers, and that they do so in a way that is more comprehensive, more instantaneous, more interactive and more opaque than earlier forms of managerial control. Since then the field has expanded in three directions. The first is empirical breadth. Research now covers ride hailing in North America, Europe, Latin America, Africa, South Asia and China; food delivery in almost every major urban market; domestic and care work platforms; and remote freelance and microtask platforms with globally distributed workforces. The second is analytical depth. Scholars have moved from cataloguing control features to examining how workers interpret, anticipate and evade them. The idea of an invisible cage captures the situation of workers who know they are being evaluated but cannot know the criteria, and who therefore engage in constant, exhausting guesswork. Related work shows how platform workers develop anticipatory practices, sharing information in forums and messaging groups in order to reconstruct the logic of the system they work under. The third direction is normative and legal, concerned with employment status, transparency rights and the design of regulation. 3.2 Control without employment A recurring theme in this literature is the mismatch between the intensity of control and the absence of the employment relationship. Studies of food delivery describe how the allocation of orders, the calculation of delivery windows, the visibility of earnings and the threat of #deactivation combine to produce a workforce that is closely governed yet formally free. Ethnographic work on couriers shows that the freedom to log off is real but costly, because the same system that permits refusal also records it. Several studies argue that this produces a distinctive form of consent. Workers frequently describe their situation in the language of entrepreneurship and #flexibility, and this language is not simply false consciousness. Flexibility has genuine value, especially for workers with caring responsibilities, students, migrants without other labour market access, and people combining several income sources. The point made by the critical literature is that flexibility is offered on terms set entirely by the platform, and that the same feature which makes the work attractive, the ability to choose when to work, is the feature which makes unpaid waiting time possible. Research on freelance and crowdwork platforms adds a further dimension. Here control is exercised less through geolocation than through reputation. A rating history is an asset that cannot be transferred between platforms, which produces a form of lock in. Workers who have accumulated a strong #reputation_score on one marketplace cannot take it elsewhere, so their bargaining position weakens over time even as their standing improves. Studies of remote gig work describe antagonism that arises not from a formal employment relation but from the subordinated agency of workers whose access to the market is mediated by a firm that is not their employer. 3.3 Political economy of platforms Parallel to the sociology of control, a political economy literature has examined the business models that make these arrangements profitable. The core observations are three. Platforms are intermediaries that capture value from transactions they do not themselves produce. They rely on network effects that tend toward concentration, which over time gives them the power of a #monopsony in local labour markets. And they treat data as an asset that is produced during the work process and appropriated without additional payment. The idea of a hidden abode of data production has been used to describe how the ordinary activity of delivering food or driving a passenger simultaneously generates a stream of information about streets, traffic, demand patterns, customer preferences and worker behaviour. This information has value that is realised in ways the worker does not participate in: it improves the allocation model, it supports the valuation of the firm on capital markets, and in some cases it is sold or licensed. Recent work on the global division of digital labour also documents how the annotation and moderation work required to build machine learning systems is itself distributed through platforms to low wage workers in the #global_south, so that the same organisational form that manages delivery riders in one country supplies the training data used to automate services in another. 3.4 Precarity, law and the classification question The legal literature has focused mainly on employment status. Courts and legislatures in the United Kingdom, Spain, the Netherlands, France, Italy, California and elsewhere have reached inconsistent conclusions about whether platform workers are employees, workers of an intermediate category, or genuinely self employed. The European Union has moved toward a presumption of employment where certain indicators of control are present, alongside transparency obligations concerning automated decision making. Legal scholars have summarised the resulting position with the observation that the boss is now an algorithm, and have argued that the substantive test of subordination should follow the reality of control rather than the label on the contract. The sociological point that follows is that #misclassification is not a defect of the model but a load bearing element of it. If platform firms were required to treat their workers as employees, the entire cost structure would change: waiting time would have to be paid, vehicles or allowances provided, social contributions made, dismissal justified. Estimates of the size of this saving vary by jurisdiction, but every serious study finds it to be substantial. 3.5 The gap this article addresses Three gaps emerge from this review. First, the literatures on control, on political economy and on law rarely speak to one another within a single framework. Sociologists describe control; political economists describe business models; lawyers describe classification. What is missing is an account that shows how these three are aspects of one thing. Second, the concept of surplus value is often invoked in passing and rarely applied with precision. It is common to read that platforms exploit workers. It is less common to read a careful statement of which form of surplus value is being extracted, through which mechanism, and with what limits. Third, the empirical literature has grown fast enough that an integrative synthesis is now both possible and necessary. This article attempts that synthesis. 4. Method 4.1 Design This is a conceptual and integrative study. It does not present new primary data. It develops a theoretical framework and tests its plausibility against the published empirical record. This design is appropriate when a field has accumulated many empirical findings that have not yet been organised by a common explanatory logic. 4.2 Corpus The evidence base consists of peer reviewed journal articles, scholarly monographs and edited volumes, and institutional research reports published mainly between 2020 and 2025. Materials were identified through database searching using combinations of the terms algorithmic management, platform work, gig economy, labour process, surplus value, digital labour and algorithmic control, and through backward and forward citation tracing from a set of anchor papers. Inclusion criteria were as follows. The study had to concern work mediated by a digital labour platform. It had to report empirical findings, offer a substantive theoretical contribution, or provide systematic descriptive data. It had to be available in English. Purely technical papers on recommendation systems, and legal commentaries without an empirical or theoretical component, were excluded except where they were needed to establish the legal position. The final corpus covers ride hailing, food and grocery delivery, domestic and care work, remote freelancing, microtasking and data annotation, across North America, Western and Southern Europe, Latin America, Sub Saharan Africa, South and East Asia and the Gulf. The geographical spread matters, because a theory of value extraction that only holds in high income economies would be a poor theory. 4.3 Analytic strategy The analysis proceeds by thematic coding against a theoretically derived scheme. Each study was read for evidence relating to five categories drawn from Section 2: the extension of unpaid time, the intensification of effort, the externalisation of costs, the appropriation of data, and the variability of pay. A sixth inductive category, worker counter conduct, emerged during coding and was retained because it bears directly on the limits of the model. Two safeguards were applied. Where studies disagreed, the disagreement is reported rather than resolved by preference. Where a claim concerns magnitude rather than mechanism, it is presented as indicative rather than definitive, because platform firms do not publish the data that would allow precise quantification. 4.4 Limitations of the design The study inherits the biases of its sources. The empirical literature over samples urban food delivery and ride hailing in a small number of countries, over samples male workers in some sectors and female workers in others, and under samples care platforms, cleaning platforms and the very large populations engaged in low paid annotation work. It also depends heavily on worker accounts, because platform firms rarely grant researchers access to their systems. These limitations are discussed further in Section 8. 5. Analysis 5.1 The working day without a working day: unpaid time and absolute surplus value The most direct route to more surplus value is more time. The historical struggle over the length of the working day produced the eight hour day, the weekend, paid rest breaks, overtime premiums and rules on standby and on call time. Every one of those institutions rests on a single legal fact: that an employer must pay for time during which the worker is at the employer's disposal. Platform work removes that fact. Because the worker is not an employee, no one is obliged to pay for availability. What remains is a category of time that is neither leisure nor paid work. The worker is logged in, waiting, ready, constrained in her movements, unable to do anything else, and receiving nothing. Empirical studies converge on the significance of this #waiting_time. Surveys of platform workers in Europe consistently find that a substantial share of the time spent working on a platform is unpaid, once searching for tasks, waiting for allocations, travelling to pick up points, communicating with clients, dealing with problems and maintaining profiles are counted. Studies of remote freelancing find that unpaid effort spent bidding for projects, writing proposals, negotiating scope and revising deliverables consumes a large fraction of total working hours. Research on food delivery describes couriers waiting near restaurant clusters, not because they wish to, but because the allocation model rewards proximity, so that unpaid positioning becomes a condition of paid work. Three mechanisms convert availability into a productive resource for the platform without converting it into a cost. The first is oversupply. Because entry is open, platforms can maintain a #reserve_army_of_labour of registered workers far larger than the number needed at any moment. Oversupply guarantees short customer waiting times, which is the service the platform sells. The cost of that service is paid in the idle hours of workers, who compete for a volume of tasks that is deliberately smaller than the volume of available labour. The second is the transfer of #dead_time. In a conventional firm, a gap in demand is a cost. The bus driver on a quiet route is still paid. The kitchen porter during a slow hour is still paid. Managers therefore have a strong incentive to smooth demand and to schedule tightly, and where they cannot, the firm absorbs the loss. On a platform, the gap is simply not paid. The incentive to smooth demand weakens, because idleness has been made free. The third is ambient availability. Applications encourage workers to remain logged in through streak bonuses, quest structures, notifications and quotas that must be completed within a window. Studies of gamified incentives find that these devices reliably extend logged in hours, and that workers describe an inability to stop, because stopping forfeits a bonus that has already been partly earned. The consequence is that the effective working day expands even when the paid working day does not. This is #absolute_surplus_value in a new form. The nineteenth century factory owner lengthened the day by keeping the gates locked. The platform lengthens the day by making a large part of it invisible. A further point deserves emphasis. #Multi_apping, the practice of running several platforms at once, is often presented as evidence of worker autonomy and market power. It is better understood as an adaptation to unpaid time. Workers run three applications because no single application supplies enough paid work to fill the hours they must be available anyway. Multi apping does not abolish unpaid waiting. It allows the worker to offer the same block of availability to several buyers, none of whom pays for it. The competition between platforms is thus resolved at the worker's expense. 5.2 Dynamic pricing, routing and gamification: the production of relative surplus value If unpaid time expands the working day, a second family of mechanisms compresses effort within it. Dynamic pricing. #Surge_pricing is publicly justified as a market clearing device. When demand exceeds supply, the price rises, which attracts more workers and rations demand. Whatever its merits as an allocation device, its function in the labour process is different. It is an instrument for the temporal and spatial concentration of labour. It moves workers to the times and places where tasks are dense, and density is precisely the condition under which the ratio of paid to unpaid time improves. A driver in a surge zone completes more trips per hour. Each trip may be better paid, but the gaps between trips shrink, so the quantity of work performed per hour of availability rises. The asymmetry lies in who controls the multiplier. The worker cannot verify the demand conditions that generate it, cannot know how long it will last, and cannot predict whether arriving in the zone will still yield the advertised rate. Several studies document the experience of chasing a surge that disappears on arrival. From the platform's point of view this is not a failure. The worker has still moved to the location where she is most useful, and the movement itself was unpaid. Routing and sequencing. Assignment systems increasingly batch tasks, so that a courier collects two or three orders on one journey, or a driver receives the next request before completing the current one. Batching is a productivity gain in the classical sense: more output per unit of labour time. The question is who captures it. Where pay is set per task and the task rate does not rise in proportion to the additional complexity, the productivity gain accrues to the platform. The courier works harder, delivers more, and earns the same or less per hour. This is the textbook definition of #relative_surplus_value: the same or greater output with a smaller share flowing to labour. Gamification. Quests, streaks, tiers, badges, progress bars and levels are widely documented. Their function is to convert an external command into an internal goal. A supervisor who ordered a courier to complete twelve deliveries before the end of a shift would be exercising managerial authority, with all the legal consequences that follow. A progress bar showing that the courier is two deliveries away from a bonus achieves a similar result without issuing any order at all. Research on this design pattern shows that it raises #labour_intensity while preserving the appearance of voluntary choice. Ratings and acceptance metrics. Customer ratings and #acceptance_rate thresholds operate as a distributed disciplinary apparatus. The platform does not need to supervise directly, because every customer is a supervisor and every declined task is recorded. The classical problem of monitoring, which historically required foremen and therefore cost money, has been solved by externalising it to customers and to software. Supervision has become nearly free, and a cost that once reduced surplus value now barely exists. The cumulative effect. These mechanisms interact. Dynamic pricing determines where the worker is. Routing determines what she does. Gamification determines how long she stays. Ratings determine whether she may continue. Each of them, taken alone, could be described as a coordination tool. Taken together they constitute a system that continuously raises the ratio of productive to unproductive time and the density of effort within productive time, which is exactly what a system for extracting relative surplus value is supposed to do. The theoretical significance should be stated carefully. In classical accounts, relative surplus value comes primarily from technological change that raises productivity in the sectors producing workers' consumption goods, thereby lowering the value of labour power across society. Platform intensification is narrower. It raises the intensity of a particular labour process without necessarily lowering the value of labour power in general. In this respect it sits closer to what the classical literature treats as intensification, an increase in the expenditure of labour within a given time, which functions like an extension of the day even though the clock does not move. The categories do not map perfectly. The point is that the platform pursues both directions of expansion at once, and that its technical apparatus is designed to do so. 5.3 The disappearing firm: externalised costs and the profit rate The third mechanism is not about time or effort at all. It is about the balance sheet. A conventional transport company owns vehicles, maintains them, insures them, fuels them, and employs the people who drive them. All of this is #constant_capital and wage cost. It appears on the balance sheet and it depresses the rate of profit relative to the surplus produced. A platform in the same market owns almost none of it. The vehicle belongs to the driver. The fuel is bought by the driver. The maintenance is arranged by the driver. Insurance, in many jurisdictions, is the driver's problem. The phone, the data plan, the thermal bag, the bicycle, the battery, the helmet, and in some markets the uniform, are all financed out of the worker's earnings. The magnitude of this transfer is easy to underestimate. Studies of courier earnings that subtract the cost of the vehicle, its depreciation, its maintenance and its energy consistently find that headline earnings figures overstate net income substantially, and that in some markets net hourly earnings fall below the applicable minimum wage once costs are deducted. Research on platform work in low and middle income countries finds the same pattern, sometimes more severely, because workers there frequently finance vehicles through credit and carry the debt themselves. Three consequences follow. The first is that the platform's capital outlay per unit of output is extraordinarily low. It carries the code, the servers, the brand and the marketing spend. This is why platform firms can scale into new cities at a speed that no asset owning competitor can match, and it is a large part of what makes them attractive to investors. The second is that risk moves downward. Accident risk, breakdown risk, illness risk, theft risk and the risk of a slow week all sit with the worker. Studies of courier safety describe workers riding in dangerous conditions because refusing to work in the rain means not eating, and because injury brings no sick pay. #Risk_transfer of this kind is a real transfer of value, because bearing risk is costly, and the party that bears it is effectively providing an unpriced service to the party that does not. The third is that the state absorbs part of the cost. Where platform earnings fall below subsistence, workers rely on in work benefits, public healthcare, family support and informal credit. The costs of reproducing the platform workforce are therefore socialised, while the surplus produced by that workforce is privatised. This is a subsidy, even though it never appears in any budget line as such. Taken together, these three effects mean that a large part of what appears as platform profitability is not a productivity achievement at all. It is an accounting achievement, produced by moving items from one side of the ledger to the other. 5.4 Data as a second circuit of appropriation Every task performed on a platform produces two outputs. The first is the service the customer buys: a completed ride, a delivered meal, an edited document. The second is a record. The record includes the route taken, the time spent at each stage, the acceptance or refusal of the offer, the rating received, the pattern of hours worked, the responsiveness to incentives and much else. This second output is not paid for. It is also not incidental. It is used to train and refine the models that allocate work, estimate delivery times, price tasks, predict which workers will accept which offers, and identify which workers are likely to leave. In this sense the labour process is simultaneously a process of #data_extraction, and the worker is producing, without additional payment, the very instrument that will be used to manage her more tightly tomorrow. Three implications matter for a theory of value. First, this is not the same as the classical case of the worker producing machinery that is later used against her. In the classical case, the machinery is a commodity produced by other workers and purchased by the capitalist. Here the instrument of control is produced as a free by product of the work it will later govern. The cost of producing the means of control is close to zero, and it is borne, in the form of effort and attention, by the workers themselves. Second, the value of the data is realised in a different circuit from the value of the service. The service is sold to a customer for a fee, from which the platform takes a commission. The data has no direct buyer in most cases, but it raises the efficiency of the allocation system, lowers the cost of matching, and supports the firm's valuation on financial markets. The literature on #platform_rent describes this as a form of value capture that does not correspond to any new production, and the description fits. Third, the same architecture is being used to produce the training material for automation. Annotation, labelling, content moderation and verification work is routed through microtask platforms to workers, often in lower income countries, who are paid per unit and who frequently have no idea what the finished system will be used for. The relation here is unusually direct. Workers are paid a piece wage to produce the inputs for systems that may later reduce the demand for the very labour they perform. This is not a paradox invented by critics. It is a straightforward consequence of a model in which the labour process and the data production process are the same process. 5.5 Personalised pay and the modern piece wage Pay on most platforms is not a fixed rate. It varies with distance, time, demand, weather, the size of the order, promotional campaigns, and, in a growing number of documented cases, with characteristics of the individual worker. The consequences for the wage relation are significant. Under a conventional piece rate, all workers doing the same task at the same time are paid the same amount. The rate is public within the workplace, which makes collective comparison possible and therefore makes collective bargaining possible. Under a personalised rate, the same task at the same moment may be worth different amounts to different workers, and no worker can see what the others are paid. #Wage_opacity of this kind destroys the informational basis of collective action. A worker who cannot know whether her rate is normal cannot know whether it is unfair. Legal scholarship has described this as #algorithmic_wage_discrimination and has argued that it functions as a form of price discrimination applied to labour rather than to goods. The firm attempts to estimate the reservation price of each worker, the minimum amount at which she will accept a given task, and to pay close to that amount. Where the estimate is accurate, the entire consumer surplus of the labour transaction, so to speak, accrues to the platform. This has three effects on surplus value. The first is direct. Paying each worker close to her individual reservation price minimises the total wage bill for a given quantity of labour, which mechanically raises the rate of exploitation. The second is indirect but larger. Variable pay makes earnings unpredictable, and unpredictable earnings force workers to work longer in order to reach a target income. The literature on target earning behaviour among drivers is contested, but the general observation is robust: when hourly income is uncertain, workers extend their hours to insure against bad days. Uncertainty is therefore not a side effect of dynamic pay. It is a productive feature of it. The third is political. Because pay is individualised and opaque, workers experience low earnings as personal failure rather than as a structural condition. Several ethnographies report workers blaming themselves for declining income, assuming that they have been penalised for some infraction they cannot identify. The system produces #self_exploitation not by deceiving workers about the rules, but by refusing to tell them what the rules are. 5.6 The limits of control: opacity, counter conduct and organisation An account that presented algorithmic management as total would be wrong, and it would also be poor sociology. Control is never complete, because labour is never fully determinate. The empirical record shows a rich repertoire of #worker_resistance. Workers reverse engineer the system. Drivers coordinate mass log offs in an area to trigger surge pricing. Couriers learn which restaurants are slow and refuse those orders in ways that do not register as refusals. Freelancers share information about which clients are abusive. Groups on messaging applications circulate advice about how to keep a rating above the threshold and how to appeal a deactivation. Studies of these practices describe a genuine collective intelligence developing outside and against the platform, and some scholars argue that the same digital tools that enable control also lower the cost of coordination among workers. Workers also organise. Despite the absence of a common workplace, a common employer and, in most jurisdictions, the legal right to bargain collectively, platform workers have mounted strikes, log offs, boycotts, litigation and public campaigns in dozens of countries. Comparative research finds that platform worker protest is a global phenomenon rather than a feature of a few rich democracies, and that food delivery couriers in particular have shown a striking capacity for spontaneous mobilisation. The very features that isolate workers, the mobile phone and the messaging group, also connect them. These findings matter for the theory of value in two ways. First, they show that the platform model has an internal contradiction. To manage labour at a distance, the platform must issue information: offers, prices, ratings, quotas, incentives. Every piece of information issued is also a piece of information disclosed. Workers use that disclosure to reconstruct the model, and once reconstructed, the model can be gamed. Platforms respond by increasing #opacity, but opacity has costs. A worker who cannot understand the incentive cannot respond to it. There is a limit beyond which secrecy destroys the very behavioural steering it was meant to protect. Second, they show that the rate of exploitation is not simply a technical parameter. It is the outcome of a struggle, and the outcome is not settled. Where couriers have organised successfully, and where courts have imposed employment status, the mechanisms described in Sections 5.1 to 5.5 have been constrained. Waiting time has been paid. Minimum earnings guarantees have been introduced. Deactivation has been made appealable. None of this abolishes the extraction of surplus value, but it changes its terms, which is precisely what the history of labour regulation has always done. 6. Discussion 6.1 A unified account The six mechanisms analysed above are usually studied separately. Placed side by side, they describe a single system with a single logic. The platform buys labour power without admitting that it is doing so. It obtains a quantity of available labour time far in excess of the time it pays for, which is a strategy of absolute surplus value. It uses pricing, routing and gamified incentives to raise the density of effort within the time it does pay for, which is a strategy of relative surplus value. It refuses to buy the means of production, transferring the cost of vehicles, energy and insurance to the worker, which raises the profit rate without any change in productivity. It appropriates, as a free by product, the data required to make the first three mechanisms more precise. It pays a variable and opaque piece rate that both minimises the wage bill and destroys the informational basis of comparison. And it manages the resulting discontent through the same channels it uses to manage the work. The unifying element is not technology as such. It is the dissolution of the employment relationship combined with the retention of managerial power. Platforms have separated the two things that were joined in the twentieth century contract of employment: the obligation of the employer and the authority of the employer. They have kept the authority and discarded the obligation. 6.2 What is new and what is not It would be an error to treat this as unprecedented. Piece work, subcontracting, day labour, home working and dependent self employment are old. The dock worker waiting at the gate for the foreman to point at him was performing unpaid availability a century ago. What is new is scale, precision and speed. Scale, because a single firm can now coordinate hundreds of thousands of nominally independent producers across dozens of countries. Precision, because pay and task allocation can be adjusted per person, per minute. Speed, because the parameters of the labour process can be changed overnight by a software update, with no negotiation, no notice period, and often no announcement. The classical categories therefore need extension rather than replacement. Two extensions are proposed here. The first is the idea of continuously adjustable exploitation. In an industrial setting the rate of exploitation was set by relatively durable institutions: the shift, the piece rate, the collective agreement, the pace of the line. Changing it required a confrontation. On a platform it can be changed silently and continuously, and it can be changed differently for different workers. The rate of exploitation becomes a control variable rather than a settled outcome. The second is the idea of costless supervision. Marx assumed that the work of superintendence was a real cost that had to be met out of surplus value. The manager, the foreman and the timekeeper had to be paid. Platforms have driven this cost close to zero by distributing supervision across customers, sensors and code. This is a genuine structural change, and it explains why platform firms can operate with astonishingly low ratios of staff to workers. 6.3 Implications for labour process theory Three implications follow for the wider field. First, control and value must be studied together. A description of algorithmic control that does not ask what the control is for produces an inventory of features rather than an explanation. Conversely, a political economy of platforms that does not examine the labour process cannot explain where the value it traces actually comes from. Second, the legal form of the contract belongs inside the analysis of the labour process, not outside it as background. #Employment_status is not a legal wrapper around an economic reality. It is one of the instruments through which the economic reality is produced. Third, the concept of #autonomy needs re examination. Platform workers have real autonomy over some things, such as when to log on, and none at all over others, such as what they are paid. The literature sometimes treats this as a paradox. It is better described as a division: autonomy is granted precisely in those domains where its exercise transfers cost and risk to the worker, and withheld in those domains where it would transfer power. 7. Policy Implications The analysis suggests four directions for policy, each of which follows from one of the mechanisms identified above. Pay for availability. If unpaid logged in time is the principal source of absolute surplus value in platform work, then any effective regulation must define and price that time. Rules that guarantee a minimum hourly income only during the performance of a task leave the mechanism intact. Regulation should attach to time at the platform's disposal, not only to time spent completing a task. Transparency as a bargaining resource, not only a consumer right. Existing transparency obligations often give individual workers a right to an explanation. Individual explanation is weak. What changes the balance of power is collective disclosure: publication of the distribution of pay rates, of acceptance thresholds, of deactivation rates and of the criteria used in allocation. #Algorithmic_accountability that stops at the individual leaves #information_asymmetry, and therefore #wage_opacity, essentially undisturbed. A presumption of employment tied to the reality of control. Where a firm sets the price, allocates the work, monitors performance and may exclude the worker from the market, the substance of subordination is present regardless of the contract's wording. Legal reforms that establish a rebuttable presumption on this basis address the classification question at its root, and directly attack the externalisation of costs described in Section 5.3. Collective rights for non employees. Competition law in several jurisdictions has historically treated self employed workers who bargain collectively as a cartel. Reform on this point is a precondition for any of the other measures to be durable, because rights that are not defended by an organisation tend not to survive. A word of caution is appropriate. Regulation raises the cost of the model, and platforms respond. Documented responses include withdrawal from markets, reclassification of workers into new intermediate categories, use of intermediary fleet companies that absorb the legal liability, and migration of work to jurisdictions with weaker enforcement. Any serious policy programme must anticipate these responses rather than be surprised by them. 8. Limitations and Future Research Four limitations should be acknowledged. First, the study relies on secondary sources. Platform firms do not disclose the workings of their allocation and pricing systems, and independent researchers have very limited access. Claims about the internal logic of these systems are therefore inferences from worker experience and from occasional disclosures in litigation, not direct observations. Second, quantification is weak. The framework predicts that the rate of exploitation is high and adjustable, but it does not permit a precise measurement of it. Constructing defensible estimates of unpaid time as a share of total working time, across sectors and countries, is an important task for future research. Third, the empirical base is geographically uneven. The great majority of published studies concern Europe, North America and China. Platform work in the Gulf, in South East Asia, in Latin America and across Africa is growing quickly and is under researched relative to its size. Because labour market conditions in these settings differ sharply, particularly with respect to migrant status and the availability of alternative employment, the mechanisms described here may operate with different intensity. Fourth, sectoral coverage is uneven. Ride hailing and food delivery dominate the literature. Care, cleaning and domestic work platforms, which employ large numbers of women and are structurally different because the work is performed inside private homes, are studied far less. So is the very large population engaged in data annotation. Four research directions follow. Comparative measurement of unpaid time. Study of the effects of regulation where it has been introduced, treating jurisdictional variation as a natural experiment. Investigation of #gendered_labour in care and cleaning platforms. And analysis of the relationship between platform work and the automation it helps to build. 9. Conclusion This article set out to explain the gig economy as a system of value extraction rather than as a technological curiosity. The conclusion is that the classical distinction between absolute and relative surplus value remains the sharpest available tool for the job, provided that it is extended to account for what is genuinely new. What is new is not that workers are exploited. What is new is that the two classical strategies for increasing exploitation have been fused into a single technical apparatus, and that this apparatus operates in the absence of the legal relationship that historically made both strategies contestable. The application on the worker's phone lengthens her day by leaving part of it unpaid, intensifies her effort by pricing and routing her movements, transfers the cost of her tools and her risks onto her, harvests the record of her work as a free input, pays her a rate she cannot see, and manages her discontent through the same interface. It does all of this while describing her as an entrepreneur. The practical implication is that reforms which address only one mechanism will fail. Paying a minimum wage during tasks does nothing about unpaid waiting. Transparency for individuals does nothing about collective invisibility. Reclassification without collective rights produces employees who cannot bargain. The mechanisms are integrated, and so must the response be. The theoretical implication is more modest but perhaps more durable. A body of thought developed to explain the textile mills of the nineteenth century turns out to describe, with very little adjustment, the delivery rider waiting outside a restaurant in a city that did not exist when that body of thought was written. That fact should be of interest to anyone who wants to understand what work is becoming, and it should temper the assumption that a new technology necessarily produces a new social relation. Sometimes it simply produces an old one, executed faster. #gig_economy #platform_capitalism #algorithmic_management #surplus_value #labour_process_theory #digital_labour #precarious_work #sociology_of_work #labour_studies #political_economy #ride_hailing #food_delivery #crowdwork #labour_rights #future_of_work References Aloisi, A. (2022). Platform work in Europe: Lessons learned, legal developments and challenges ahead. European Labour Law Journal, 13(1), 4 to 29. Altenried, M. (2022). The Digital Factory: The Human Labor of Automation. Chicago: University of Chicago Press. Anwar, M. A., and Graham, M. (2021). Between a rock and a hard place: Freedom, flexibility, precarity and vulnerability in the gig economy in Africa. Competition and Change, 25(2), 237 to 258. Barratt, T., Goods, C., and Veen, A. (2020). I am my own boss: Active intermediation and entrepreneurial worker agency in the Australian gig economy. Environment and Planning A: Economy and Space, 52(8), 1643 to 1661. Bessa, I., Joyce, S., Neumann, D., Stuart, M., Trappmann, V., and Umney, C. (2022). A Global Analysis of Worker Protest in Digital Labour Platforms. ILO Working Paper 70. Geneva: International Labour Organization. Bonini, T., and Trere, E. (2024). Algorithms of Resistance: The Everyday Fight against Platform Power. Cambridge, MA: MIT Press. Bucher, E. L., Schou, P. K., and Waldkirch, M. (2021). Pacifying the algorithm: Anticipatory control practices of platform workers. Organization, 28(1), 44 to 67. Cini, L. (2023). Resisting algorithmic control: Understanding the rise and variety of platform worker mobilisations. New Technology, Work and Employment, 38(1), 125 to 144. Cini, L., and Goldmann, B. (2021). The worker capabilities approach: Insights from worker mobilizations in Italian logistics and food delivery. Work, Employment and Society, 35(5), 948 to 967. Cutolo, D., and Kenney, M. (2021). Platform dependent entrepreneurs: Power asymmetries, risks, and strategies in the platform economy. Academy of Management Perspectives, 35(4), 584 to 605. De Stefano, V., and Aloisi, A. (2022). Your Boss Is an Algorithm: Artificial Intelligence, Platform Work and Labour. Oxford: Hart Publishing. Delfanti, A. (2021). The Warehouse: Workers and Robots at Amazon. London: Pluto Press. Dubal, V. (2023). On algorithmic wage discrimination. Columbia Law Review, 123(7), 1929 to 1992. Duggan, J., Sherman, U., Carbery, R., and McDonnell, A. (2020). Algorithmic management and app work in the gig economy: A research agenda for employment relations and HRM. Human Resource Management Journal, 30(1), 114 to 132. Englert, S., Woodcock, J., and Cant, C. (2020). Digital workerism: Technology, platforms, and the circulation of workers struggles. tripleC: Communication, Capitalism and Critique, 18(1), 132 to 145. Ferrari, F., and Graham, M. (2021). Fissures in algorithmic power: Platforms, code, and contestation. Cultural Studies, 35(4 to 5), 814 to 832. Fuchs, C. (2021). Digital Capitalism: Media, Communication and Society, Volume Three. London: Routledge. Graham, M., and Ferrari, F. (Eds.) (2022). Digital Work in the Planetary Market. Cambridge, MA: MIT Press. Gregory, K. (2021). My life is more valuable than this: Understanding risk among on demand food couriers in Edinburgh. Work, Employment and Society, 35(2), 316 to 331. Harvey, D. (2023). A Companion to Marx's Grundrisse. London: Verso. Heiland, H. (2021). Controlling space, controlling labour? Contested space in food delivery gig work. New Technology, Work and Employment, 36(1), 1 to 16. Heiland, H. (2022). Neither timeless, nor placeless: Control of food delivery gig work via place based working time regimes. Human Relations, 75(9), 1824 to 1848. Howson, K., Ustek Spilda, F., Bertolini, A., Heeks, R., Ferrari, F., Katta, S., et al. (2022). Stripping back the mask: Working conditions on digital labour platforms during the COVID 19 pandemic. International Labour Review, 161(3), 413 to 440. International Labour Organization (2021). World Employment and Social Outlook 2021: The Role of Digital Labour Platforms in Transforming the World of Work. Geneva: ILO. Jarrahi, M. H., Newlands, G., Lee, M. K., Wolf, C. T., Kinder, E., and Sutherland, W. (2021). Algorithmic management in a work context. Big Data and Society, 8(2), 1 to 14. Joyce, S., Neumann, D., Trappmann, V., and Umney, C. (2020). A Global Struggle: Worker Protest in the Platform Economy. ETUI Policy Brief 2. Brussels: European Trade Union Institute. Karanovic, J., Berends, H., and Engel, Y. (2021). Regulated dependence: Platform workers responses to new forms of organizing. Journal of Management Studies, 58(4), 1070 to 1106. Kellogg, K. C., Valentine, M. A., and Christin, A. (2020). Algorithms at work: The new contested terrain of control. Academy of Management Annals, 14(1), 366 to 410. Kenney, M., Bearson, D., and Zysman, J. (2021). The platform economy matures: Measuring pervasiveness and exploring power. Socio Economic Review, 19(4), 1451 to 1483. Kougiannou, N. K., and Mendonca, P. (2021). Breaking the managerial silencing of worker voice in platform capitalism: The rise of a food courier network. British Journal of Management, 32(3), 744 to 759. Lei, Y. W. (2021). Delivering solidarity: Platform architecture and collective contention in China's platform economy. American Sociological Review, 86(2), 279 to 309. Mendonca, P., Kougiannou, N. K., and Clark, I. (2023). Informalization in gig food delivery in the UK: The case of hyper flexible and precarious work. Industrial Relations: A Journal of Economy and Society, 62(1), 60 to 77. Milkman, R., Elliott Negri, L., Griesbach, K., and Reich, A. (2021). Gender, class, and the gig economy: The case of platform based food delivery. Critical Sociology, 47(3), 357 to 372. Miceli, M., and Posada, J. (2022). The data production dispositif. Proceedings of the ACM on Human Computer Interaction, 6(CSCW2), 1 to 37. Moore, P. V., and Woodcock, J. (Eds.) (2021). Augmented Exploitation: Artificial Intelligence, Automation and Work. London: Pluto Press. Muldoon, J. (2022). Platform Socialism: How to Reclaim Our Digital Future from Big Tech. London: Pluto Press. Muldoon, J., and Raekstad, P. (2023). Algorithmic domination in the gig economy. European Journal of Political Theory, 22(4), 587 to 607. Newlands, G. (2021). Algorithmic surveillance in the gig economy: The organization of work through Lefebvrian conceived space. Organization Studies, 42(5), 719 to 737. Piasna, A., Zwysen, W., and Drahokoupil, J. (2022). The Platform Economy in Europe: Results from the Second ETUI Internet and Platform Work Survey. ETUI Working Paper 2022.05. Brussels: European Trade Union Institute. Pulignano, V., Piasna, A., Domecka, M., Muszynski, K., and Vermeerbergen, L. (2021). Does It Pay to Work? Unpaid Labour in the Platform Economy. ETUI Policy Brief 15. Brussels: European Trade Union Institute. Purcell, C., and Brook, P. (2022). At least I am my own boss: Explaining consent, coercion and resistance in platform work. Work, Employment and Society, 36(3), 391 to 406. Rahman, H. A. (2021). The invisible cage: Workers reactivity to opaque algorithmic evaluations. Administrative Science Quarterly, 66(4), 945 to 988. Rani, U., and Furrer, M. (2021). Digital labour platforms and new forms of flexible work in developing countries: Algorithmic management of work and workers. Competition and Change, 25(2), 212 to 236. Ravenelle, A. J. (2023). Side Hustle Safety Net: How Vulnerable Workers Survive Precarious Times. Oakland: University of California Press. Schor, J. B. (2020). After the Gig: How the Sharing Economy Got Hijacked and How to Win It Back. Oakland: University of California Press. Stark, D., and Pais, I. (2020). Algorithmic management in the platform economy. Sociologica, 14(3), 47 to 72. Sun, P., Chen, J. Y., and Rani, U. (2023). From flexible labour to sticky labour: A tracking study of workers in the food delivery platform economy of China. Work, Employment and Society, 37(2), 412 to 431. Ticona, J. (2022). Left to Our Own Devices: Coping with Insecure Work in a Digital Age. New York: Oxford University Press. Tubaro, P., Casilli, A. A., and Coville, M. (2020). The trainer, the verifier, the imitator: Three ways in which human platform workers support artificial intelligence. Big Data and Society, 7(1), 1 to 12. Vallas, S., and Schor, J. B. (2020). What do platforms do? Understanding the gig economy. Annual Review of Sociology, 46, 273 to 294. Vallas, S. P., Johnston, H., and Mommadova, Y. (2022). Prime suspect: Mechanisms of labor control at Amazon's warehouses. Work and Occupations, 49(4), 421 to 456. van Doorn, N., and Badger, A. (2020). Platform capitalism's hidden abode: Producing data assets in the gig economy. Antipode, 52(5), 1475 to 1495. van Doorn, N., Ferrari, F., and Graham, M. (2023). Migration and migrant labour in the gig economy: An intervention. Work, Employment and Society, 37(4), 1099 to 1111. Vandaele, K. (2021). Collective resistance and organizational creativity amongst Europe's platform workers: A new power in the labour movement? In J. Haidar and M. Keune (Eds.), Work and Labour Relations in Global Platform Capitalism (pp. 206 to 235). Cheltenham: Edward Elgar. Veen, A., Barratt, T., and Goods, C. (2020). Platform capital's app etite for control: A labour process analysis of food delivery workers in Australia. Work, Employment and Society, 34(3), 388 to 406. Waldkirch, M., Bucher, E., Schou, P. K., and Grunwald, E. (2021). Controlled by the algorithm, coached by the crowd: How HRM activities take shape on digital work platforms in the gig economy. International Journal of Human Resource Management, 32(12), 2643 to 2682. Wells, K. J., Attoh, K., and Cullen, D. (2023). Disrupting D.C.: The Rise of Uber and the Fall of the City. Princeton: Princeton University Press. Wood, A. J., and Lehdonvirta, V. (2021). Antagonism beyond employment: How the subordinated agency of labour platforms generates conflict in the remote gig economy. Socio Economic Review, 19(4), 1369 to 1396. Wood, A. J., and Lehdonvirta, V. (2023). Platforms disrupting reputation: Precarity and recognition struggles in the remote gig economy. Sociology, 57(5), 1035 to 1053. Woodcock, J. (2021). The Fight Against Platform Capitalism: An Inquiry into the Global Struggles of the Gig Economy. London: University of Westminster Press.

This article revisits Karl Marx's concept of the metabolic rift and asks whether it still explains the ecological condition of the twenty-first century. The dominant policy language treats #global_warming, soil exhaustion, water stress and biodiversity loss as "market failures": unpriced side effects that better accounting, carbon pricing or greener technology can correct. Using a critical political economy approach and an integrative review of recent Earth system science, ecological economics and Marxian ecology (2020 to 2025), this study argues the opposite. It argues that #ecological_degradation is not a malfunction of the capitalist system but an expression of how the system is built to work. The paper reconstructs the rift in three linked forms: a material rift in the circulation of nutrients, carbon and water; a social and spatial rift produced by unequal exchange between the Global North and the Global South; and a temporal rift in which future ecological conditions are systematically discounted to protect present accumulation. Evidence is drawn from planetary boundary assessments, global carbon budget data, nitrogen and phosphorus flow studies, decoupling and rebound literature, and empirical work on ecologically unequal exchange. The analysis shows that absolute decoupling of economic growth from resource use at the speed required by climate targets has no empirical precedent, that efficiency gains are repeatedly absorbed by expanded output, and that the burdens of extraction are displaced onto peripheral regions rather than removed. The paper concludes that #climate_policy which leaves the growth imperative untouched can slow the rift but cannot close it, and it outlines what a metabolic repair programme, grounded in #degrowth, #agroecology, public planning and reparative justice, would have to involve. The contribution is theoretical and synthetic rather than empirical: it offers students and researchers a coherent framework for reading contemporary environmental data through the category of #social_metabolism. Keywords: metabolic rift; social metabolism; ecological Marxism; climate crisis; degrowth; ecologically unequal exchange; planetary boundaries; political economy of nature 1. Introduction Every year the scientific record becomes clearer and the policy response remains recognisably the same. Global mean surface temperature keeps climbing, the annual carbon budget consistent with a 1.5 degree Celsius pathway keeps shrinking, and international negotiations keep producing pledges whose implementation gap is measured in gigatonnes. Ripple and colleagues, reviewing planetary vital signs, described the situation as one of entry into uncharted territory, with most indicators moving in the wrong direction despite three decades of climate governance (Ripple et al., 2023, 2024). At the same time, assessments of the planetary boundaries framework concluded that six of nine boundaries have already been transgressed, including those for climate change, biosphere integrity, land system change, freshwater change, biogeochemical flows and novel entities (Richardson et al., 2023). Faced with this record, mainstream #environmental_economics offers a familiar diagnosis. Pollution and depletion are described as externalities: costs that fall outside the price mechanism because property rights are incomplete or information is imperfect. The prescription follows from the diagnosis. Price the externality, internalise the cost, and the market will allocate resources efficiently again. Carbon taxes, emissions trading, natural capital accounting, biodiversity offsets and green finance all rest on this logic. The underlying assumption is that the crisis is a #market_failure, an accident of institutional design that can be repaired without touching the design of the economy itself. This article challenges that assumption. It argues that #global_warming and #resource_depletion are better understood as structural products of an economic system whose reproduction depends on continuous expansion. The theoretical instrument for making this argument is not new. It was developed in the nineteenth century by Karl Marx, who drew on the agricultural chemistry of Justus von Liebig to describe how capitalist agriculture broke the circulation of nutrients between countryside and city. Marx wrote in the third volume of Capital that large landed property produces conditions that provoke an irreparable rift in the interdependent process of social metabolism, a metabolism prescribed by the natural laws of life itself (Marx, 1981). John Bellamy Foster later reconstructed this passage into a full theory, giving it the name by which it is now known: the theory of the #metabolic_rift (Foster, 1999). The concept has since become one of the most productive frameworks in environmental sociology and ecological Marxism (Foster, 2022; Saito, 2023). Yet it is often introduced to students as a historical curiosity, a nineteenth century observation about guano and manure that happens to sound prophetic. This article treats it instead as an operational analytical tool. The claim being tested is straightforward. If the metabolic rift theory is correct, then ecological breakdown should not appear as a random or correctable byproduct of production but as a patterned, repeated and expanding feature of accumulation, one that reappears in new form each time it is technologically or geographically displaced. 1.1 Research problem The research problem can be stated in one sentence. Three decades of #green_capitalism have produced significant technological progress, considerable regulatory activity, falling costs for renewable energy, and no reduction in global material throughput or in absolute atmospheric concentrations of carbon dioxide. Global material extraction continues to rise, and global emissions have not entered sustained decline at the rate required by any credible 1.5 degree pathway (Friedlingstein et al., 2023). If the market failure hypothesis were adequate, the sustained application of market corrections over thirty years ought to have produced measurable convergence toward ecological limits. It has not. This gap between remedy and result is the empirical anomaly that motivates the study. 1.2 Research questions The article addresses four questions. What exactly did Marx mean by metabolic rift, and how has the concept been developed, contested and extended in the recent literature? Can contemporary ecological disruptions, specifically the carbon rift, the nutrient rift, the hydrological rift and the biodiversity rift, be coherently described using this framework? Is there empirical support for the claim that these disruptions are structurally necessary to capital accumulation rather than accidental to it? What follows for policy and praxis if the structural interpretation is accepted? 1.3 Aim and contribution The aim is to provide a rigorous, readable synthesis that connects a body of critical theory with current environmental science. The intended contribution is threefold. First, the paper consolidates the recent revival of Marxian ecology, especially the work associated with the new reading of Marx's notebooks, into a single accessible framework. Second, it tests that framework against specific and recent empirical literatures rather than asserting it. Third, it draws out the implications for #climate_governance, arguing that policy which treats growth as a constant and ecology as a variable has the relationship backwards. 1.4 Structure of the article Section 2 reviews the intellectual history and current state of metabolic rift theory, including its main critics. Section 3 sets out the conceptual framework, distinguishing material, spatial and temporal dimensions of the rift. Section 4 explains the method. Section 5 presents the analysis across six domains. Section 6 discusses counterarguments. Section 7 considers implications for policy and practice. Section 8 states limitations and a research agenda, and Section 9 concludes. 2. Literature Review 2.1 The chemical origins of a social concept The concept of metabolism entered nineteenth century science through physiology and agricultural chemistry. Justus von Liebig, whose work on plant nutrition established that soil fertility depends on the replacement of specific mineral nutrients, became increasingly critical of what he called the robbery system of British agriculture. Intensive farming exported nitrogen, phosphorus and potassium in the form of food and fibre to distant cities, where those nutrients accumulated as waste and pollution instead of returning to the fields that produced them. The soil was being mined rather than cultivated. This was not a metaphor but a measurable depletion of #soil_fertility, and it produced a frantic international search for replacement inputs: Peruvian guano, Chilean nitrates, and later bones imported from European battlefields and catacombs. Marx read Liebig closely. The recent scholarly reconstruction of his excerpt notebooks, published through the Marx Engels Gesamtausgabe project, shows that his engagement with natural science was not decorative but central to the later development of his critique of political economy (Saito, 2023). Marx concluded that capitalist agriculture achieves short term productivity gains by undermining the conditions of long term fertility, and that the same logic applies to the worker. Progress in capitalist agriculture, he wrote, is progress in the art not only of robbing the worker but of robbing the soil. 2.2 Foster's reconstruction and the first generation of rift theory For most of the twentieth century this argument was largely ignored, and Marxism was frequently characterised as productivist and indifferent to nature. Foster's intervention reversed that reading by showing that a systematic ecological argument was embedded in Capital itself, organised around the category of #social_metabolism, the material exchange between human society and the rest of nature mediated by labour (Foster, 1999). In this reconstruction the rift has a specific meaning. It is not simply pollution or damage. It is a break in a cycle: a flow of matter that was previously circular becomes linear, running from extraction through use to accumulation as waste. Foster and his collaborators subsequently extended the analysis from soil to a general theory of ecological crisis, identifying rifts in the carbon cycle, the nitrogen cycle, the hydrological cycle and marine ecosystems (Foster, 2022; Foster and Clark, 2020). The theory of #unequal_ecological_exchange was developed alongside it, showing that rifts are frequently not repaired but relocated, exported to peripheral regions through trade and colonial relations. 2.3 The second wave: Saito and the ecological reading of Marx The most significant recent development is the work of Kohei Saito, who used the MEGA notebooks to argue that Marx's ecological thinking deepened substantially after 1868, moving away from any residual faith in the progressive character of capitalist development. On Saito's reading, the later Marx came to see #capital_accumulation as fundamentally incompatible with the sustainable reproduction of natural conditions, and became interested in communal, non-capitalist forms of land management. Saito draws the conclusion that a coherent Marxian ecology today points toward what he calls #degrowth_communism: a steady state, needs oriented economy under democratic social control (Saito, 2023, 2024). This reading has been influential and contested in equal measure. It has, however, decisively shifted the terrain of debate. The question is no longer whether Marx had an ecology, but what kind of ecological politics follows from it. 2.4 The critique from world ecology The most serious internal challenge comes from Jason W. Moore, who argues that rift theory retains an implicit Cartesian dualism. By speaking of a rift between society and nature, the theory risks treating the two as separate entities that were once joined and have now been torn apart. Moore proposes instead the language of metabolic shift and of capitalism as a world-ecology, a way of organising nature rather than something acting upon it from outside (Moore, 2017). Capitalism, in this account, does not merely deplete nature; it produces natures, and its historical dynamism has depended on appropriating what Moore calls the unpaid work of cheap food, cheap energy, cheap raw materials and cheap labour. Defenders of rift theory reply that the criticism misreads the concept. The rift is not between society and nature as substances but within a metabolic process in which human labour is one moment. Marx's own definition of labour as the metabolic interaction between human beings and nature already rules out the dualism Moore attributes to the theory (Foster, 2022). The debate is not fully resolved, and this article does not attempt to settle it. What matters for present purposes is that both positions agree on the central empirical claim: capital's ecological contradictions are internal to its logic and not external to it. 2.5 Adjacent literatures Three further literatures inform the analysis. The first is #ecological_economics, and particularly the empirical work on decoupling. Systematic reviews find no evidence of absolute, global, permanent decoupling of resource use from GDP growth at anything close to the rate required for climate stabilisation (Vaden et al., 2020; Hickel et al., 2022a). The rebound literature reinforces this, showing that economy wide rebound effects absorb a substantial share of energy efficiency gains (Brockway et al., 2021). The second is the literature on #ecologically_unequal_exchange. Quantitative studies find large net flows of embodied land, labour, energy and materials from the Global South to the Global North, with the drain running to trillions of dollars in value terms (Dorninger et al., 2021; Hickel et al., 2021, 2022b). This is the empirical face of the displaced rift. The third is the critical scholarship on green capitalism, which examines why market instruments underperform. Buller (2022) analyses the conceptual failures of carbon markets, offsets and natural capital valuation, and Christophers (2024) shows that renewable electricity has struggled to attract private capital at the required scale because its profitability, not its cost, is the binding constraint. Both accounts point to the same conclusion: the problem is not that the market has failed to see nature, but that it can only see nature in the form of #profitability. 2.6 Research gap Despite this productive literature, three gaps remain. First, the theoretical and the empirical strands rarely meet: rift theory tends to be argued philosophically, while the decoupling and unequal exchange evidence tends to be presented without a unifying theory of why the patterns recur. Second, the temporal dimension of the rift, especially in relation to overshoot and future carbon removal, has received far less attention than the material and spatial dimensions. Third, teaching-oriented syntheses that make the framework usable for students are scarce. This article addresses all three. 3. Conceptual Framework The framework used here treats the metabolic rift as a single logic that expresses itself in three registers. Separating them is analytically useful because policy responses tend to address one while intensifying another. 3.1 The material rift The material rift is the original formulation. It refers to any interruption in the cyclical movement of matter and energy through an ecosystem caused by the organisation of production. Its signature is the conversion of a loop into a line. Nutrients that once returned to the soil accumulate as #eutrophication in rivers and coastal dead zones. Carbon that was locked in geological storage for hundreds of millions of years is transferred, in the space of two centuries, into the atmosphere. Water withdrawn from aquifers at rates exceeding recharge becomes soil moisture deficit and then, eventually, salinised land. In each case something that was a #biogeochemical_cycle becomes a throughput. The material rift has a characteristic tempo. It is slow to appear and fast to compound, because ecological systems absorb disturbance until a threshold is crossed. This mismatch between the tempo of ecological damage and the tempo of quarterly profit is one reason markets systematically underdetect it. 3.2 The spatial rift The second register concerns geography. Capital rarely repairs a rift when it can move it. When British soils were exhausted, fertility was imported as guano from Peru, extracted under conditions of indentured Chinese labour. When European forests were depleted, timber came from the colonies. The pattern persists. The nutrient rift of European livestock production is today closed, partially and temporarily, by soy grown on converted savanna and forest in South America. The carbon rift of Northern consumption is partially concealed by locating emissions intensive manufacturing in Asia. The mineral demands of the energy transition are met by #extractivism in the Democratic Republic of the Congo, Chile, Bolivia and Indonesia. Displacement is not repair. It transfers the rift to a place where its costs are borne by populations with less political capacity to refuse. This is what recent scholarship names #climate_coloniality: the persistence of colonial relations of extraction and disposability inside the ostensibly universal project of climate action (Sultana, 2022). 3.3 The temporal rift The third register is the least discussed and, in the present conjuncture, the most consequential. A temporal rift occurs when the reproduction of present accumulation depends on borrowing from ecological futures that cannot be repaid. Discounting in cost benefit analysis performs this operation formally, assigning lower present value to damages that occur later. Integrated assessment models perform it structurally, when they permit near term emissions to exceed the budget on the assumption that #negative_emissions technologies will remove the excess in the second half of the century. Malm and Carton (2024) describe the result as a politics of overshoot: a tacit acceptance that limits will be breached, coupled with a promise that they will be un-breached later. The promise is largely unfunded, unbuilt and physically speculative at the scale assumed (Carton et al., 2021). The temporal rift is therefore the mechanism by which the material rift is not solved but deferred, and deferral has a direction: it moves the burden onto people who are not yet able to object. 3.4 The accumulation engine Underlying all three registers is a single mechanism. In the circuit of capital, money is advanced to purchase means of production and labour power, commodities are produced, and they are sold for a larger sum of money. The purpose of production is not the satisfaction of need but the valorisation of value. Because the surplus must be reinvested to avoid devalorisation by competitors, the circuit has no natural stopping point. Growth is not a preference within the system; it is the condition of the system's survival. This is why a #growth_imperative cannot be legislated away while the underlying relations remain intact, and it is the precise point at which Marxian ecology parts company with mainstream #sustainability discourse. The framework generates a testable proposition. If ecological damage were an externality, we would expect its correction to be compatible with continued accumulation, and we would expect corrective instruments to produce absolute reductions in throughput. If it is a rift in the sense described here, we would expect corrective instruments to produce relative improvements in efficiency that are absorbed by expanded scale, and we would expect the damage to reappear in displaced form. Section 5 examines which expectation the evidence supports. 4. Methodology 4.1 Research design This is a conceptual and theoretical study using the method of critical political economy combined with an integrative literature review. It does not generate new primary data. Its purpose is explanatory: to assess whether an existing theoretical framework accounts for an established body of empirical findings better than the rival framework that currently dominates policy. The design follows three steps. First, the theory is reconstructed from primary and secondary sources. Second, testable expectations are derived from it, as set out in Section 3.4. Third, those expectations are compared against recent empirical literature in Earth system science, ecological economics, agrarian studies and political ecology. 4.2 Selection of sources The evidence base was assembled through structured searching of peer reviewed literature published mainly between 2020 and 2025, supplemented by a small number of foundational texts that remain indispensable, notably Marx's Capital and Foster's original reconstruction. Priority was given to systematic reviews, global assessments and quantitative studies with international coverage. Where a claim is contested in the literature, the contest is reported rather than concealed. Three inclusion criteria were applied. Sources had to be peer reviewed or published by an established academic press; they had to address global or macro-regional patterns rather than single case anecdotes; and, for empirical claims, they had to report methods sufficient to allow assessment of their limits. 4.3 Analytical strategy Analysis proceeds by pattern matching. For each ecological domain, the study asks three questions. Does the domain display the signature of a rift, that is, the conversion of a cycle into a linear throughput? Has the disruption been displaced rather than closed? And have market based corrective instruments produced absolute reductions or only relative efficiency gains? Convergent answers across independent domains strengthen the structural interpretation; divergent answers would weaken it. 4.4 Limitations of the method Three limitations should be stated at the outset. First, theoretical synthesis cannot establish causation in the statistical sense; it can establish coherence and explanatory adequacy. Second, source selection inevitably involves judgement, and a scholar working from a different tradition would weight the same literature differently. Third, the argument concerns a system as a whole, and system level claims are difficult to falsify cleanly because no controlled counterfactual exists. These limits are revisited in Section 8. 5. Analysis 5.1 The carbon rift: fossil capital and the atmospheric commons Coal, oil and gas are the compressed remains of biological productivity accumulated over geological time. Combustion returns that carbon to the atmosphere in an instant of geological reckoning. The slow biological and geological processes that could return it to storage operate over tens of thousands of years. This is the clearest possible example of a #carbon_rift: a closed long cycle opened into a one directional flow. What makes the carbon rift structural rather than accidental is the role fossil energy plays in the accumulation process. Fossil fuels are not simply one energy source among several. They are storable, portable, spatially concentrated and controllable, which allowed production to be relocated to where labour was cheapest and most disciplined rather than to where water power happened to flow. Fossil energy was, from the beginning, a technology of #social_power as much as of thermodynamics. The empirical record is consistent with the structural reading in several respects. First, energy transitions have historically been energy additions. Global renewable capacity has expanded rapidly, and yet absolute fossil fuel consumption has not fallen correspondingly, because total energy demand has grown alongside it. Efficiency has improved almost continuously for two centuries without producing a decline in aggregate throughput, a pattern anticipated by the Jevons paradox and confirmed by contemporary reviews of #rebound_effects, which find that economy wide rebound absorbs a large proportion of expected savings (Brockway et al., 2021). Second, the pattern of investment reveals the priority. Fossil capital is a stock of assets whose valuation depends on the assumption that reserves will be burned. Semieniuk and colleagues estimated that the write down of stranded fossil assets under a rapid transition would impose losses concentrated among investors in advanced economies, largely transmitted through financial intermediaries and pension funds (Semieniuk et al., 2022). The system therefore has an enormous, internally rational, distributed interest in delay. That interest is not a conspiracy; it is a balance sheet. Third, the industry's own knowledge of the problem has been documented. Analysis of internal projections shows that fossil fuel companies produced accurate climate modelling decades before they publicly conceded the science, and that historical accountability can be traced through internal corporate records (Supran et al., 2023; Bonneuil et al., 2021). What is significant for the argument here is not the moral failing but the systemic rationality of the behaviour. A firm that had voluntarily written down its reserves in 1985 would have been outcompeted, acquired or replaced. #Competition enforces the rift. Fourth, the carbon rift is deeply unequal in its production. Analysis of global carbon inequality shows that the top ten per cent of emitters account for around half of global emissions, while the bottom half account for a small fraction (Chancel, 2022). This matters analytically because it undermines the demographic explanation of the crisis. The rift tracks #accumulation, not population. 5.2 The soil and nutrient rift: agriculture as extraction The original rift is also the most persistent. Industrial agriculture separates the site of nutrient uptake from the site of nutrient deposition and then bridges the gap with synthetic inputs. The Haber-Bosch process, which fixes atmospheric nitrogen using fossil energy, is the technological patch that allowed the nutrient rift to be sustained through the twentieth century. It did not close the rift; it converted a fertility problem into a pollution problem and an energy dependency. The results are visible in the biogeochemical record. Reactive nitrogen now flows through terrestrial and aquatic systems at rates far above pre-industrial levels, and analysis of regional nitrogen boundaries indicates that agricultural nitrogen pollution exceeds safe limits across large parts of Europe, South Asia and East Asia, even where the global average appears less alarming (Schulte-Uebbing et al., 2022). Excess nitrogen and phosphorus drive #eutrophication, hypoxic zones and freshwater contamination. Phosphorus presents an additional problem of finitude: it has no synthetic substitute, is geologically concentrated in a handful of countries, and is currently used in a linear rather than a circular manner (Brownlie et al., 2022). A civilisation dependent on mined phosphate is running a metabolic deficit that no price signal can abolish. The soil itself is being degraded. Global assessments of land and water resources report widespread degradation of agricultural land, with erosion rates that exceed soil formation rates over large areas (FAO, 2022). Food systems as a whole are responsible for roughly a third of global anthropogenic greenhouse gas emissions, with a substantial share coming from land use change and livestock (Crippa et al., 2021). Two structural features explain the persistence of the pattern. The first is #corporate_concentration. A small number of firms dominate seeds, agrochemicals, machinery, grain trading and processing. Concentration reduces the range of viable agronomic choices available to farmers, because inputs, credit, insurance and market access are bundled into a package that presupposes high input monoculture (Clapp, 2021). A farmer who wishes to close nutrient loops on the farm faces a supply chain designed for the opposite. The second is the #value_form itself. Soil organic matter, mycorrhizal networks and pollinator populations do not appear on a balance sheet. They are, in the language of the classical economists, free gifts of nature. Because they are free, their depletion registers as cost saving rather than as capital consumption. The accounting system is not merely blind to the rift; it rewards it. Agroecological alternatives exist and are well documented. Diversified systems that integrate livestock and cropping, use cover crops and green manures, and rely on ecological rather than chemical pest control can maintain yields while rebuilding soil carbon and reducing external inputs (Anderson et al., 2021; van der Ploeg, 2021). Their limited uptake is not a scientific problem. It is a problem of land tenure, credit, subsidy structures and market power, which is to say a problem of political economy. 5.3 The oceanic and hydrological rift The rift extends into water. Roughly a quarter of anthropogenic carbon dioxide has been absorbed by the oceans, which has lowered surface pH measurably. #Ocean_acidification is an unusually clean illustration of the concept, because it is not a side effect of a poorly designed market but a direct chemical consequence of the volume of emissions, unaffected by whether those emissions were priced. No conceivable carbon market would have altered the chemistry, only the quantity. Marine fisheries display a parallel logic. Under competitive accumulation, a fish stock is a resource whose future value is uncertain and whose present capture is certain. Fleets expand, capital is sunk into vessels and processing capacity, and the resulting overcapacity generates political pressure to maintain quotas above sustainable levels. The rift here is between the reproductive cycle of the stock and the turnover cycle of the capital invested in catching it. Freshwater shows the same structure with an added spatial twist. Assessments now identify a distinct planetary boundary for green water, the moisture held in soil and vegetation, and conclude that it has been transgressed (Wang-Erlandsson et al., 2022). Groundwater depletion in major agricultural basins is driven by the export orientation of production: aquifers in water scarce regions are drawn down to grow water intensive crops for distant markets. The water is not consumed locally; it is embodied in commodities and shipped away, a form of #virtual_water transfer that constitutes hydrological unequal exchange. Biodiversity loss completes the picture. Reviews of extinction evidence conclude that current rates are far above background rates and consistent with an ongoing sixth mass extinction event, driven overwhelmingly by habitat conversion and land use change (Cowie et al., 2022). Habitat conversion is, in economic terms, the transformation of land that produces no revenue into land that produces revenue. #Biodiversity is destroyed not despite the rationality of the system but because of it. 5.4 The displaced rift: ecologically unequal exchange and green extractivism If the rift were simply a technical problem, we would expect wealthy economies, having the most advanced technology and the strictest regulation, to have closed it. Territorial data appear to support this: several high income countries report falling domestic emissions and improving local environmental quality. Consumption based accounting tells a different story. Once trade is taken into account, much of the apparent improvement dissolves. Quantitative work on #ecologically_unequal_exchange finds that high income countries appropriate large net quantities of embodied land, materials, energy and labour from low and middle income countries. Dorninger and colleagues documented systematic asymmetries in the physical composition of trade, with poorer countries exporting resource intensive goods and importing value intensive ones (Dorninger et al., 2021). Hickel and colleagues quantified the resulting drain in monetary terms, finding transfers on a scale that dwarfs official development assistance flowing in the opposite direction (Hickel et al., 2021, 2022b). Fair share assessments of resource use likewise attribute the overwhelming majority of overshoot beyond safe planetary limits to high income nations (Hickel et al., 2022c). The pattern reproduces itself inside the #energy_transition. Decarbonisation is materially intensive. Wind turbines, solar arrays, transmission infrastructure and batteries require copper, nickel, cobalt, lithium, and rare earth elements in large volumes. Projected mineral demand under rapid transition scenarios implies a substantial expansion of extraction, and the geography of that expansion falls disproportionately on the Global South and on indigenous territories (Riofrancos, 2023; Sovacool, 2021). Modelling of the energy requirements of the transition itself indicates that building the new system consumes energy and generates emissions that are not trivial, particularly if the transition is attempted while total energy demand continues to grow (Slamersak et al., 2022). This is not an argument against renewable energy. It is an argument about the relationship between the transition and the growth imperative. A transition that replaces fossil infrastructure within a stable or contracting energy demand envelope is materially manageable. A transition that must simultaneously power continuous expansion becomes a new frontier of extraction, and reproduces the rift in a different mineral register. The distinction is decisive, and it is invisible to any framework that treats growth as a background constant. The equity dimension has been analysed under the heading of #climate_coloniality. Sultana argues that dominant climate discourse universalises a "we" that erases who has caused the crisis, who suffers from it, and whose land is required to solve it (Sultana, 2022). Hickel and Slamersak show that many mainstream mitigation scenarios assume the Global South will continue supplying cheap resources and will forgo the energy use required for decent living standards, thereby encoding colonial inequalities into the models themselves (Hickel and Slamersak, 2022). Reparative frameworks respond by treating climate justice as a question of historical accountability rather than charitable assistance (Taiwo, 2022). 5.5 The temporal rift: overshoot, offsets and the promise of future removal The most sophisticated contemporary form of rift displacement is temporal. It works through three mechanisms. The first is #offsetting. Carbon offsets permit an emitter to continue emitting on the condition that an equivalent quantity is avoided or absorbed elsewhere. The equivalence is theoretically elegant and physically dubious. Fossil carbon released from geological storage remains in the active carbon cycle for millennia. Carbon stored in a plantation is vulnerable to fire, disease, harvest and policy reversal within decades. Treating the two as fungible is an accounting convention, not a chemical fact, and critical scholarship has termed this problem the undoing of equivalence (Carton et al., 2021). Investigations of offset quality have repeatedly found that a large share of credited reductions are not additional. The second is #net_zero framing. A net target permits gross emissions to continue provided they are balanced by removals. Because removals are cheap on paper and expensive in reality, the framing systematically encourages the deferral of hard abatement in favour of hypothetical future subtraction. Malm and Carton describe the resulting settlement as a managed surrender: overshoot is not a failure of the plan but has quietly become the plan (Malm and Carton, 2024). The third is #discounting. Standard cost benefit analysis applies a positive discount rate to future damages, which means that harms occurring in the twenty second century are assigned a present value approaching zero. The discount rate is presented as a technical parameter. It is in fact an ethical statement about whose suffering counts, converted into arithmetic. Together these mechanisms constitute a #temporal_rift that mirrors the spatial one. Where the spatial rift exports the ecological cost to people who are far away, the temporal rift exports it to people who are not yet born. In neither case is the cost abolished. It is merely moved to a location from which it cannot be politically contested. 5.6 Why this is not a market failure The preceding sections identify a consistent pattern across six domains. It is now possible to state the theoretical claim precisely. The market failure hypothesis holds that ecological damage is an unpriced external cost, and that pricing it will restore efficient allocation. This hypothesis makes an implicit assumption that deserves to be made explicit: it assumes that there exists a price vector at which the system can both internalise ecological costs and continue to accumulate. If no such price vector exists, the language of externality collapses, because an externality that cannot be internalised without destroying the mechanism that generates it is not an externality at all. It is a #contradiction. Four bodies of evidence suggest that the required price vector does not exist. First, the #decoupling evidence. Systematic reviews of the empirical literature find cases of relative decoupling, in which resource use grows more slowly than GDP, and some cases of absolute decoupling of territorial emissions in individual wealthy countries, largely explained by outsourcing and by fuel switching. What they do not find is evidence of absolute, global, sufficiently rapid, permanent decoupling of material throughput from output (Vaden et al., 2020). The rates required to reconcile continued growth with the remaining carbon budget have no historical precedent and no demonstrated mechanism (Hickel et al., 2022a; Keysser and Lenzen, 2021). Decoupling is not impossible in principle; it is unevidenced at the required magnitude, which for policy purposes is the relevant fact. Second, the #rebound evidence. Efficiency improvements reduce the effective cost of energy services, which raises demand for them and frees expenditure for other consumption. Economy wide rebound estimates cluster around a large fraction of the technically expected savings (Brockway et al., 2021). Under competitive accumulation, efficiency is not a route to sufficiency; it is a route to expanded scale. This is not a distortion of the market. It is the market functioning correctly. Third, the #profitability constraint. Christophers demonstrates that the slow build out of renewable generation is not chiefly explained by cost, since renewables are now frequently the cheapest source of new electricity, but by the difficulty of earning stable, attractive returns in electricity markets characterised by near zero marginal cost and volatile prices (Christophers, 2024). Capital does not invest in what is cheap or necessary. It invests in what is profitable. Where the socially necessary and the profitable diverge, the market chooses profit, and no amount of information disclosure will change that ranking. Fourth, the #valuation critique. Attempts to render nature legible to markets through natural capital accounting, biodiversity offsets and carbon pricing repeatedly encounter the same difficulty: ecological relations are qualitative, place specific, non-substitutable and often non-linear, whereas exchange value is quantitative, abstract, substitutable and continuous (Buller, 2022; Corvellec et al., 2022). Rendering the former as the latter necessarily destroys the information that mattered. The measure deletes what it was meant to protect. Taken together, these findings support the structural interpretation. The rift is not a failure of the market to include nature. It is a consequence of the market including nature on its own terms, that is, as a source of surplus. The system does not malfunction when it degrades the biosphere. It functions. 6. Discussion 6.1 Engaging the ecomodernist counterargument The strongest objection to the argument advanced here comes from ecological modernisation and its more recent ecomodernist variants. The objection runs as follows. Human ingenuity has repeatedly falsified predictions of resource limits. Energy is not scarce, since the solar flux vastly exceeds human requirements. Dematerialisation is possible, and information intensive economies use fewer materials per unit of output. Therefore the correct strategy is to accelerate technological substitution, decouple aggressively, and use growth to fund the transition. This position deserves a serious response rather than dismissal, and three points can be made. The first is empirical. The claim that dematerialisation is occurring at the aggregate global level is not supported. Global material extraction has grown roughly in step with output. Apparent dematerialisation in high income economies largely reflects the geographical relocation of material intensive production, as the consumption based accounting literature demonstrates (Dorninger et al., 2021). The second is thermodynamic. Information can be dematerialised; food, housing, transport, steel and cement cannot. A substantial fraction of the material economy consists of physical requirements that shrink only within limits. The third is political. Even if a technically feasible decoupling pathway could be specified, it would still have to be implemented by institutions whose reproduction depends on expansion. Feasibility in a model is not feasibility in a political economy. The ecomodernist argument tends to assume the existence of a benevolent planner who is missing from its own account of the system. Where the ecomodernists are right, and this must be conceded, is in their insistence that technology matters and that renewable energy is indispensable. Nothing in the metabolic rift framework counsels technological pessimism. The framework claims only that technology deployed inside an accumulation imperative will be used to expand throughput, not to reduce it. 6.2 Engaging the world ecology critique Moore's objection, discussed in Section 2.4, is that rift language smuggles in dualism. Two responses are possible. The analytical response is that the rift is defined within a metabolic relation, not between two separate substances. What is torn is a process, not a boundary between society and a pristine nature that never existed. The strategic response is that the two frameworks converge where it matters. Whether one says that capitalism ruptures social metabolism or that capitalism organises nature through the appropriation of unpaid work, the political conclusion is the same: the ecological crisis cannot be resolved without transforming the relations of production. The debate is a debate among allies about the best vocabulary for a shared diagnosis. 6.3 The state socialism objection A serious objection notes that centrally planned economies of the twentieth century produced severe environmental damage, from the desiccation of the Aral Sea to industrial pollution across Eastern Europe. If socialism also produced ecological catastrophe, then capital cannot be the explanation. The objection has force and should not be evaded. Two clarifications are required. First, the historical record shows that the Soviet model reproduced the #productivism it inherited, pursuing output maximisation with the same indifference to metabolic cycles, and in some respects with fewer feedback mechanisms because independent scientific criticism and public protest were suppressed. Rift theory does not claim that only capitalism can rupture social metabolism. It claims that capitalism must, because expansion is structurally compelled rather than politically chosen. A planned economy may choose growth; a capitalist economy cannot choose not to. Second, the conclusion that follows is not that planning is irrelevant but that planning is a necessary and insufficient condition. Democratic accountability, ecological science and the absence of a growth compulsion must all be present. This is precisely the argument made by contemporary ecological socialists, who insist that #ecological_planning must be democratic and needs-based rather than output-maximising (Vettese and Pendergrass, 2022; Saito, 2024). 6.4 The Malthusian objection and why it fails A recurrent response to ecological crisis attributes it to population. The metabolic rift framework rejects this, and the emissions data support the rejection. Carbon inequality analysis shows that consumption is concentrated at the top of the global income distribution, and that the emissions of the poorest half of humanity are a small fraction of the total (Chancel, 2022). Studies of decent living standards find that universal provision of sufficient energy for health, nutrition, shelter, education and mobility is achievable at a fraction of current global energy use (Millward-Hopkins et al., 2020; Kikstra et al., 2021). The binding problem is not the number of people but the organisation of production and the distribution of throughput. Population framings, moreover, tend to direct blame toward those least responsible, which is why they have long been criticised as a form of #environmental_racism. 6.5 What would falsify the argument? Intellectual honesty requires specifying what evidence would count against the thesis. The structural interpretation would be substantially weakened if, over a sustained period, global material footprint and global greenhouse gas emissions declined absolutely and rapidly while global output continued to grow, without that decline being explained by outsourcing, recession or the temporary effects of a pandemic. It would be further weakened if market instruments alone, without regulatory quantity constraints, produced such a decline. No such evidence currently exists. If it emerged, the theory would have to be revised. That is what distinguishes a theory from a doctrine. 7. Implications: From Diagnosis to Metabolic Repair If the analysis holds, then the policy question changes. The question is no longer how to make growth green, but how to organise a decent life for eight billion people within a contracting material envelope. Six implications follow. 7.1 Sufficiency and the deliberate reduction of throughput The first implication is that #throughput must be treated as the object of policy, not as an outcome to be inferred. This is the core insight of the degrowth research programme, which is frequently misunderstood as a call for austerity or recession. It is neither. Degrowth proposes a planned, equitable reduction of energy and material use in high income economies, coupled with expansion of provisioning where needs are unmet, while maintaining or improving wellbeing (Hickel et al., 2022a; Fitzpatrick et al., 2022). Modelling work shows that low energy demand pathways make the 1.5 degree target more attainable and reduce reliance on speculative negative emissions (Keysser and Lenzen, 2021). Concretely this means scaling down destructive and low utility sectors, including fossil fuel extraction, mass advertising, planned obsolescence, private aviation and fast fashion, while scaling up care, health, education, public transport, insulation and renewable energy. This is not less economic activity across the board. It is a different composition of activity, chosen deliberately rather than accepted as an aggregate. 7.2 Public provisioning and the reduction of the profit dependency The second implication concerns the mechanism of provision. If profitability is the binding constraint on decarbonisation, then removing the profitability requirement from the sectors that must be transformed is not an ideological luxury but an engineering necessity. Public ownership of electricity generation and transmission, public housing retrofit programmes, and free or low cost public transport all convert investments that are unattractive to private capital into investments that are rational for society (Christophers, 2024). #Public_luxury and private sufficiency is a slogan, but it names a real allocation principle: collective provision of high quality goods reduces per capita throughput far more effectively than individual consumption choices. 7.3 Agroecology and the closing of nutrient loops The third implication returns to the original rift. Repair requires reconnecting nutrient cycles: reintegrating livestock and cropping, composting and returning urban organic waste to agricultural land, recovering phosphorus from wastewater, using leguminous rotations to fix nitrogen biologically, and rebuilding soil organic matter. #Agroecology is the scientific and practical body of knowledge that does this, and it is well developed (Anderson et al., 2021; van der Ploeg, 2021). Its obstacles are political. Closing loops requires proximity between production and consumption, which conflicts with global commodity trading. It requires labour, which conflicts with the substitution of capital for labour. It requires secure land tenure, which conflicts with land concentration and financialisation. #Land_reform and the defence of the commons are therefore not adjacent to ecological policy; they are ecological policy (Angus, 2023). 7.4 Reparations and the closing of the spatial rift The fourth implication is that any credible programme must address the displaced rift. This means technology transfer without extractive intellectual property regimes, debt cancellation to remove the pressure that forces indebted countries to export raw materials, financing for adaptation and loss and damage on a scale commensurate with historical responsibility, and mineral governance that respects consent and prohibits the reproduction of colonial extraction under a green label (Sultana, 2022; Taiwo, 2022). #Climate_reparations is the name for treating the ecological debt as a debt. 7.5 Democratic ecological planning The fifth implication concerns coordination. Reducing throughput while guaranteeing needs cannot be achieved by price signals alone, because prices allocate according to purchasing power rather than need. Some form of #democratic_planning is required: quantity constraints on emissions and extraction, public investment directed by deliberative processes, and sectoral plans developed with unions and affected communities (Vettese and Pendergrass, 2022). The historical failures of authoritarian planning are an argument for democratising planning, not for abandoning coordination to a market that has demonstrated it cannot perform this function. 7.6 Labour, class and the politics of transition The sixth implication is strategic. An ecological politics that presents itself as a demand for reduced consumption, addressed to a population already experiencing declining living standards, will be defeated. Huber argues that climate politics must be built through the material interests of the working class, framing decarbonisation as a struggle over who controls production and for whose benefit (Huber, 2022). #Just_transition guarantees, employment in the retrofit and care economies, and reduced working time are not concessions attached to ecological policy; they are the conditions under which such policy can be enacted at all. The rift will not be closed by consumers. It will be closed, if at all, by producers organised politically. 8. Limitations and Future Research This study has clear limits. It is theoretical rather than empirical, and it therefore inherits the evidential quality of the literature it synthesises. Where that literature is contested, notably in the estimation of economy wide rebound effects and in the modelling of low demand mitigation pathways, the conclusions here are correspondingly provisional. It also operates at a high level of abstraction. Speaking of capitalism as a single system risks obscuring real variation between financialised Anglo-American capitalism, coordinated European capitalism, state-directed Chinese development and extractive peripheral economies. Those differences affect what is politically possible, and a more granular comparative analysis would refine the argument considerably. The article treats the metabolic rift concept as broadly valid while acknowledging the world ecology critique. A fuller engagement would attempt to specify the empirical conditions under which the two frameworks yield different predictions, which they may not. Four research directions follow. First, quantitative work is needed on the material and energy requirements of a transition undertaken under contrasting demand assumptions, disaggregated by region. Second, comparative institutional analysis should examine which existing forms of collective ownership have measurably reduced throughput. Third, more rigorous evaluation of #agroecological transitions at scale is required, including labour and yield outcomes. Fourth, the temporal rift deserves systematic study: the political economy of overshoot and of carbon removal promises remains under-theorised relative to its policy importance. 9. Conclusion The argument of this article can be compressed into a single proposition. The ecological crisis is not a bug in the operating system of the modern economy. It is a feature of that system's design, and it will be reproduced by any configuration of that design, however green its interface. Marx observed that capitalist agriculture disturbed the metabolic interaction between humanity and the earth by taking fertility from the soil and failing to return it. The scale has changed beyond recognition. The structure has not. Fossil carbon is extracted from geological storage and not returned. Nitrogen and phosphorus are mined, applied, washed into water bodies and not returned. Aquifers are drawn down and not recharged. Forests are converted to pasture and cropland and not restored. Minerals are extracted from the territories of the poor to power the consumption of the rich, and the resulting waste is not recovered. In every case a cycle has been opened into a line, and in every case the line terminates in a sink that cannot absorb it. Three decades of market-based #environmental_governance have not reversed this. They have produced improved efficiency per unit of output alongside a larger number of units, cleaner air in wealthy cities alongside displaced pollution in poorer ones, falling territorial emissions in the North alongside rising embodied emissions in Northern consumption, and an accounting architecture that converts physical impossibility into balance sheet equivalence. The instruments have worked exactly as designed. It is the design that is the problem. This is not a counsel of despair, and it is not an argument against acting now with the tools available. Regulation, renewable deployment, efficiency standards and carbon pricing all reduce harm at the margin, and marginal harm reduction saves real lives. The argument is that these measures cannot be the horizon of ambition, because they operate inside a logic that will absorb their gains and continue to expand. Closing the rift requires acting on the logic itself: producing for need rather than for accumulation, reducing throughput deliberately in the places that consume most, restoring the cycles that industrial production has broken, and repaying what has been taken from those who did not take it. Marx called the rift irreparable under the conditions he was describing. He was right about those conditions. Whether it remains irreparable is not a scientific question. It is a political one, and it is still open. Hashtags #Metabolic_Rift #Ecological_Marxism #Climate_Crisis #Social_Metabolism #Degrowth #Political_Ecology #Planetary_Boundaries #Ecological_Economics #Environmental_Sociology #Capitalism_And_Nature #Green_Capitalism_Critique #Ecosocialism #Climate_Justice #Sustainability_Studies #Earth_System_Science References Anderson, C. R., Bruil, J., Chappell, M. J., Kiss, C., and Pimbert, M. P. (2021). Agroecology Now! Transformations Towards More Just and Sustainable Food Systems. Cham: Palgrave Macmillan. Angus, I. (2023). The War Against the Commons: Dispossession and Resistance in the Making of Capitalism. New York: Monthly Review Press. Bonneuil, C., Choquet, P.-L., and Franta, B. (2021). Early warnings and emerging accountability: Total's responses to global warming, 1971-2021. Global Environmental Change, 71, 102386. Brockway, P. E., Sorrell, S., Semieniuk, G., Heun, M. K., and Court, V. (2021). Energy efficiency and economy-wide rebound effects: A review of the evidence and its implications. Renewable and Sustainable Energy Reviews, 141, 110781. Brownlie, W. J., Sutton, M. A., Reay, D. S., Heal, K. V., Hermann, L., Kabbe, C., and Spears, B. M. (2022). Global actions for a sustainable phosphorus future. Nature Food, 2, 71-74. Buller, A. (2022). The Value of a Whale: On the Illusions of Green Capitalism. Manchester: Manchester University Press. Carton, W., Lund, J. F., and Dooley, K. (2021). Undoing equivalence: Rethinking carbon accounting for just carbon removal. Frontiers in Climate, 3, 664130. Chancel, L. (2022). Global carbon inequality over 1990-2019. Nature Sustainability, 5, 931-938. Christophers, B. (2024). The Price is Wrong: Why Capitalism Won't Save the Planet. London: Verso. Clapp, J. (2021). The problem with growing corporate concentration and power in the global food system. Nature Food, 2, 404-408. Corvellec, H., Stowell, A. F., and Johansson, N. (2022). Critiques of the circular economy. Journal of Industrial Ecology, 26(2), 421-432. Cowie, R. H., Bouchet, P., and Fontaine, B. (2022). The Sixth Mass Extinction: Fact, fiction or speculation? Biological Reviews, 97(2), 640-663. Crippa, M., Solazzo, E., Guizzardi, D., Monforti-Ferrario, F., Tubiello, F. N., and Leip, A. (2021). Food systems are responsible for a third of global anthropogenic GHG emissions. Nature Food, 2, 198-209. Dorninger, C., Hornborg, A., Abson, D. J., von Wehrden, H., Schaffartzik, A., Giljum, S., Engler, J.-O., Feller, R. L., Hubacek, K., and Wieland, H. (2021). Global patterns of ecologically unequal exchange: Implications for sustainability in the 21st century. Ecological Economics, 179, 106824. Fitzpatrick, N., Parrique, T., and Cosme, I. (2022). Exploring degrowth policy proposals: A systematic mapping with thematic synthesis. Journal of Cleaner Production, 365, 132764. Food and Agriculture Organization of the United Nations (2022). The State of the World's Land and Water Resources for Food and Agriculture: Systems at Breaking Point. Rome: FAO. Foster, J. B. (1999). Marx's theory of metabolic rift: Classical foundations for environmental sociology. American Journal of Sociology, 105(2), 366-405. Foster, J. B. (2022). Capitalism in the Anthropocene: Ecological Ruin or Ecological Revolution. New York: Monthly Review Press. Foster, J. B., and Clark, B. (2020). The Robbery of Nature: Capitalism and the Ecological Rift. New York: Monthly Review Press. Friedlingstein, P., O'Sullivan, M., Jones, M. W., Andrew, R. M., Bakker, D. C. E., Hauck, J., et al. (2023). Global Carbon Budget 2023. Earth System Science Data, 15(12), 5301-5369. Hickel, J., Dorninger, C., Wieland, H., and Suwandi, I. (2022b). Imperialist appropriation in the world economy: Drain from the global South through unequal exchange, 1990-2015. Global Environmental Change, 73, 102467. Hickel, J., Kallis, G., Jackson, T., O'Neill, D. W., Schor, J. B., Steinberger, J. K., Victor, P. A., and Urge-Vorsatz, D. (2022a). Degrowth can work: Here's how science can help. Nature, 612, 400-403. Hickel, J., O'Neill, D. W., Fanning, A. L., and Zoomkawala, H. (2022c). National responsibility for ecological breakdown: A fair-shares assessment of resource use, 1970-2017. The Lancet Planetary Health, 6(4), e342-e349. Hickel, J., and Slamersak, A. (2022). Existing climate mitigation scenarios perpetuate colonial inequalities. The Lancet Planetary Health, 6(7), e628-e631. Hickel, J., Sullivan, D., and Zoomkawala, H. (2021). Plunder in the post-colonial era: Quantifying drain from the Global South through unequal exchange, 1960-2018. New Political Economy, 26(6), 1030-1047. Huber, M. T. (2022). Climate Change as Class War: Building Socialism on a Warming Planet. London: Verso. Keysser, L. T., and Lenzen, M. (2021). 1.5 degrees C degrowth scenarios suggest the need for new mitigation pathways. Nature Communications, 12, 2676. Kikstra, J. S., Mastrucci, A., Min, J., Riahi, K., and Rao, N. D. (2021). Decent living gaps and energy needs around the world. Environmental Research Letters, 16(9), 095006. Malm, A., and Carton, W. (2024). Overshoot: How the World Surrendered to Climate Breakdown. London: Verso. Marx, K. (1976). Capital: A Critique of Political Economy, Volume I. Translated by B. Fowkes. London: Penguin. Marx, K. (1981). Capital: A Critique of Political Economy, Volume III. Translated by D. Fernbach. London: Penguin. Millward-Hopkins, J., Steinberger, J. K., Rao, N. D., and Oswald, Y. (2020). Providing decent living with minimum energy: A global scenario. Global Environmental Change, 65, 102168. Moore, J. W. (2017). Metabolic rift or metabolic shift? Dialectics, nature, and the world-historical method. Theory and Society, 46(4), 285-314. Richardson, K., Steffen, W., Lucht, W., Bendtsen, J., Cornell, S. E., Donges, J. F., et al. (2023). Earth beyond six of nine planetary boundaries. Science Advances, 9(37), eadh2458. Riofrancos, T. (2023). The security-sustainability nexus: Lithium onshoring in the Global North. Global Environmental Politics, 23(1), 20-41. Ripple, W. J., Wolf, C., Gregg, J. W., Levin, K., Rockstrom, J., Newsome, T. M., et al. (2023). The 2023 state of the climate report: Entering uncharted territory. BioScience, 73(12), 841-850. Ripple, W. J., Wolf, C., Gregg, J. W., Rockstrom, J., Mann, M. E., Oreskes, N., et al. (2024). The 2024 state of the climate report: Perilous times on planet Earth. BioScience, 74(12), 812-824. Saito, K. (2023). Marx in the Anthropocene: Towards the Idea of Degrowth Communism. Cambridge: Cambridge University Press. Saito, K. (2024). Slow Down: The Degrowth Manifesto. New York: Astra House. Schulte-Uebbing, L. F., Beusen, A. H. W., Bouwman, A. F., and de Vries, W. (2022). From planetary to regional boundaries for agricultural nitrogen pollution. Nature, 610, 507-512. Semieniuk, G., Holden, P. B., Mercure, J.-F., Salas, P., Pollitt, H., Jobson, K., Vercoulen, P., Chewpreecha, U., Edwards, N. R., and Vinuales, J. E. (2022). Stranded fossil-fuel assets translate to major losses for investors in advanced economies. Nature Climate Change, 12, 532-538. Slamersak, A., Kallis, G., and O'Neill, D. W. (2022). Energy requirements and carbon emissions for a low-carbon energy transition. Nature Communications, 13, 6932. Sovacool, B. K. (2021). When subterranean slavery supports sustainability transitions? Power, patriarchy, and child labor in artisanal Congolese cobalt mining. The Extractive Industries and Society, 8(1), 271-293. Sultana, F. (2022). The unbearable heaviness of climate coloniality. Political Geography, 99, 102638. Supran, G., Rahmstorf, S., and Oreskes, N. (2023). Assessing ExxonMobil's global warming projections. Science, 379(6628), eabk0063. Taiwo, O. O. (2022). Reconsidering Reparations. New York: Oxford University Press. Vaden, T., Lahde, V., Majava, A., Jarvensivu, P., Toivanen, T., Hakala, E., and Eronen, J. T. (2020). Decoupling for ecological sustainability: A categorisation and review of research literature. Environmental Science and Policy, 112, 236-244. van der Ploeg, J. D. (2021). The political economy of agroecology. The Journal of Peasant Studies, 48(2), 274-297. Vettese, T., and Pendergrass, D. (2022). Half-Earth Socialism: A Plan to Save the Future from Extinction, Climate Change and Pandemics. London: Verso. Wang-Erlandsson, L., Tobian, A., van der Ent, R. J., Fetzer, I., te Wierik, S., Porkka, M., Staal, A., Jaramillo, F., Dahlmann, H., Singh, C., Greve, P., Gerten, D., Keys, P. W., Gleeson, T., Cornell, S. E., Steffen, W., Bai, X., and Rockstrom, J. (2022). A planetary boundary for green water. Nature Reviews Earth and Environment, 3, 380-392.

This paper asks a simple question with complicated answers. Can a theory written in the 1860s explain a market that did not exist until 2009? The theory is Karl #Marx's account of #fictitious_capital, developed mainly in Volume III of Capital, where he describes paper claims that trade at prices set by expectations rather than by the value of any productive asset behind them. The market is the crypto economy: coins, tokens, stablecoins, non fungible tokens and the enormous derivatives layer built on top of them. The paper builds a five part analytical framework from Marx's categories of #interest_bearing_capital, #capitalisation and the #credit_system, and then applies that framework to five crypto sub markets using published empirical research and official financial stability reporting from 2021 to 2026. The finding is mixed, and the mixture is the contribution. Marx explains the structure of crypto crises very well: the pyramiding of claims on claims, the reflexive role of #collateral, the way #leverage turns a price fall into a chain of forced sales, and the sudden destruction of paper wealth that leaves the physical economy largely untouched. Marx explains the substance of crypto assets less well, because his fictitious capital is a discounted claim on a future income stream, while a coin such as #Bitcoin has no income stream to discount. The paper therefore proposes an extension rather than a simple application. It argues that crypto assets are best read as fictitious capital in which the fictional revenue has itself been abolished, leaving only the #capitalisation mechanism, and that this pure form makes the instability described by Marx faster, more visible and more frequent rather than less. The paper closes with implications for #regulation, for teaching political economy, and for future empirical work. Keywords: fictitious capital, Marxian political economy, cryptocurrency, decentralised finance, non fungible tokens, financial crises, leverage, financialisation 1. Introduction Between November 2021 and November 2022, roughly two trillion United States dollars of recorded crypto market value disappeared. Almost nothing else disappeared with it. No factories closed because a token lost its peg. No harvests failed because a lending protocol was drained. A handful of firms collapsed, some of them spectacularly, and a very large number of individual investors lost savings they could not afford to lose, but the physical productive base of the world economy carried on more or less as before. That combination, an enormous loss of recorded wealth with almost no loss of productive capacity, is exactly what #Marx predicted when he wrote about the destruction of paper claims during a crisis. He argued that when such claims are wiped out, the real means of production remain intact, and what has vanished is a set of entitlements to future income rather than the income producing machinery itself. This paper takes that observation seriously and asks whether the Marxian category of #fictitious_capital, developed in the third volume of Capital, can be used as a working analytical tool for the digital asset economy rather than as a rhetorical flourish. The question matters for three reasons. First, it matters for economic theory. Mainstream finance has produced a large and technically sophisticated literature on crypto asset pricing, market microstructure, manipulation and contagion. That literature is very good at describing what happens. It is less comfortable explaining why a whole asset class whose cash flows are close to zero can be capitalised at trillions of dollars for years at a time, and why the collapses look so similar to one another. Marxian #political_economy starts from precisely that puzzle, because for Marx the gap between paper valuation and productive substance is not an anomaly to be corrected but a structural feature of a developed #credit_system. Second, it matters for #financial_stability policy. Official bodies have moved in a few years from treating crypto as a curiosity to treating it as a potential source of #systemic_risk, particularly through #stablecoins, leverage and the growing links between crypto firms and regulated finance. If the instability of the sector is driven by something structural rather than by bad luck or bad actors, then policy that focuses only on fraud and disclosure will keep arriving late. Third, it matters for students. The crypto economy is the first financial system that most young people encounter directly, often before they have opened a savings account. It is unusually transparent, because most transactions are recorded publicly, and unusually fast, because its boom and bust cycles run in months rather than decades. This makes it an almost ideal teaching laboratory for the classical questions of political economy: what is value, where does profit come from, and why does credit periodically destroy itself. 1.1 Research problem The problem this paper addresses can be stated plainly. Marx defined fictitious capital as a tradable claim whose price is arrived at by discounting an expected future revenue at the prevailing rate of interest. Shares, bonds and bank credit all fit that definition. A share is a claim on future profit, a government bond is a claim on future tax revenue, and both are priced by capitalisation. But a Bitcoin is not a claim on anything. It pays no dividend, promises no coupon and entitles its holder to no share of anyone's #surplus_value. If fictitious capital requires a fictional revenue stream, then strictly speaking Bitcoin is not fictitious capital at all. It might be something else entirely: a collectible, a commodity money, a speculative token, or simply a bet. Yet the behaviour of crypto markets in a crisis is uncannily close to the behaviour Marx described. That mismatch, a category that seems to fit the dynamics but not the definition, is the research problem. 1.2 Research questions The paper is organised around four questions. What, precisely, did Marx mean by fictitious capital, and which of its features are essential rather than incidental to his argument? To what extent do #cryptocurrencies, #stablecoins, #DeFi lending protocols, #NFTs and crypto #derivatives satisfy those essential features? Can the Marxian account of credit driven crisis explain the observed pattern of crypto crashes between 2020 and 2026, including the Terra and FTX episodes? Where does the Marxian framework fail, and what modification would be needed to make it work? 1.3 Contribution The paper makes three contributions. Theoretically, it distinguishes between the substance of fictitious capital, which is the fictional revenue stream, and the mechanism of fictitious capital, which is capitalisation plus credit plus tradability, and it argues that crypto reproduces the mechanism while abolishing the substance. Analytically, it builds a five criterion test that can be applied to any digital asset to judge how far it belongs to the category. Empirically, it organises the existing quantitative literature on crypto crashes around a Marxian crisis sequence, showing that the observed sequence of events in 2022 matches the theoretical sequence closely. 1.4 Structure Section 2 reviews the literature. Section 3 sets out the theoretical framework. Section 4 explains the method. Sections 5 to 9 present the analysis of the five sub markets. Section 10 draws the crisis sequence together. Section 11 discusses limitations, and Section 12 concludes. 2. Literature Review 2.1 The Marxian literature on fictitious capital Marx introduced fictitious capital in the discussion of the #credit_system in Volume III of Capital, and it has always been one of the more slippery categories in his work. He used it to describe several different things: bank notes, public debt, share certificates, and the duplication of the same underlying capital in several sets of books at once. The core idea is that once a stream of income can be sold, it acquires a price, and that price is calculated by dividing the expected income by the rate of interest. The result is a paper asset that circulates, that can be bought on credit, that can be pledged as security for further borrowing, and whose price moves for reasons that have little to do with the machinery or the workers who actually generate the income. Marxian scholarship on this category was relatively thin for most of the twentieth century, and interest revived only with the financialisation debates after 2008. Recent theoretical work has clarified the definition considerably. Alves (2023) argues, against much of the earlier Marxian literature, that fictitious capital does not represent any real capital at all, and that it is best understood as the channel through which #interest_bearing_capital comes to dominate other forms of capital. On this reading the fiction is not that the paper is worthless, since it clearly has a price and confers real command over resources, but that it presents itself as capital in its own right, as if it were an independent source of income rather than a claim on income produced somewhere else. This reading is important for the present paper, because it removes the requirement that a title of fictitious capital must have a real productive asset standing behind it. A second strand looks at how digital technology reshapes the form of value. Fuchs (2022) analyses the platform economy in Marxian terms and stresses the difficulty of locating productive labour in digital systems. Harvey (2023), in his reading of the Grundrisse, emphasises the way credit and speculation accelerate the circuit of capital until the circuit begins to consume itself. Lapavitsas and colleagues (2023) place financial profit at the centre of contemporary capitalism and argue that finance now extracts income directly from households rather than only from productive enterprise. Varoufakis (2023) and Durand (2024) both argue, in different ways, that the dominant form of income in the digital economy is #rent rather than profit, and that platform owners collect tribute from those who use their infrastructure. That argument matters here because several crypto assets look far more like rent bearing property than like claims on industrial profit. The most direct precursor to the present paper is Comyn (2026), who theorises cryptocurrency as fictitious money, a social form anchored in the general equivalent that helps circulate and valorise new digital assets. Comyn's argument, developed through the monetary theory of Elson and de Brunhoff, is that crypto should not be dismissed as a non money simply because it circulates badly, but should be read as a part of the money form that reveals money's own contradictions. The present paper is complementary. Where Comyn asks what crypto tells us about the #money_form, this paper asks what crypto tells us about the capital form, and specifically about capitalisation. 2.2 The financial economics literature on crypto assets The mainstream literature is now very large. Fang and colleagues (2022) provide a comprehensive survey of crypto trading research covering market conditions, pricing, prediction and portfolio construction. Baur and Dimpfl (2021) study the #volatility of Bitcoin and conclude that its extreme price variability undermines both its use as a #medium_of_exchange and its use as a #store_of_value, which is an awkward result for an asset whose main claimed function is to be money. Makarov and Schoar (2022) survey crypto and #DeFi and document the concentration of holdings, the dominance of speculative motives and the limited real world usage of the underlying networks. On manipulation, Cong and colleagues (2023) estimate that #wash_trading accounts for a very large share of reported volume on unregulated exchanges, with the proportion rising sharply on smaller venues. Their work matters here because reported market capitalisation, the headline number that supposedly measures crypto wealth, is calculated from prices set in markets where a large share of the trading is not real trading at all. The literature on #NFTs reaches similar conclusions, documenting extensive self dealing and artificially generated volume in the leading marketplaces. On #DeFi, the analytical work of Aramonte, Huang and Schrimpf (2021) is central. They identify what they call the decentralisation illusion: because some level of governance is unavoidable, and because consensus mechanisms and token ownership both tend to concentrate power, DeFi is far more centralised in practice than in principle. They also list the vulnerabilities that make the sector fragile, namely high #leverage, liquidity mismatch, built in interconnectedness and the absence of any shock absorbing institution comparable to a bank with a central bank behind it. A companion study by Aramonte, Doerr, Huang and Schrimpf (2022) shows that DeFi lending relies on #overcollateralisation because the system cannot screen borrowers, and that this reliance makes the credit supply strongly #procyclical: collateral values rise, borrowing capacity rises with them, and the process runs in reverse on the way down. Auer and colleagues (2024) describe the technical architecture of DeFi in detail, and a later Bank for International Settlements study (Auer, Borio and others, 2025) sets out the financial stability implications systematically. On the crises themselves, Briola and colleagues (2023) provide a forensic account of the #TerraUSD failure, tracing the mechanics of the depeg and the death spiral between the stablecoin and its governance token. Liu, Makarov and Schoar (2023) analyse the same event as a classic run, showing how the Anchor protocol's unsustainable yield created the conditions for a coordinated exit. Badev and Watsky (2023) trace the ripple effects of the Terra collapse across interconnected DeFi protocols. Official assessments by the Financial Stability Board (2022, 2023) and the International Monetary Fund (2022) reach broadly similar conclusions about leverage, opacity, interconnection and the absence of shock absorbers. 2.3 The gap Two literatures therefore exist, and they do not talk to each other. The Marxian literature has the conceptual apparatus for explaining why paper claims periodically self destruct, but it has produced relatively little detailed engagement with the mechanics of crypto markets, and where it has engaged, it has usually done so at the level of #money rather than capital. The financial economics literature has excellent data and precise mechanisms but no theory of why the asset class exists at all, and it tends to treat each crash as a governance failure or an information failure rather than as an expression of something structural. This paper sits in the gap. It uses the mechanisms documented by the empirical literature as evidence, and it uses the Marxian categories as the explanatory frame. 3. Theoretical Framework 3.1 From value to price to capitalisation Marx's argument runs in stages, and it is worth setting them out slowly, because the argument about crypto depends on getting the sequence right. The starting point is the distinction between #use_value and #exchange_value. A thing has use value if it satisfies a want. It has exchange value if it can be traded for other things in a definite proportion. Marx argued that in a commodity producing society the magnitude of exchange value is regulated, on average and over time, by the socially necessary labour time required to reproduce the commodity. This is the #labour_theory_of_value, and it is the most contested part of his system. For present purposes the important point is not whether the theory is correct in every detail but what follows from it, namely that value has a substance located in #production, and that everything happening in #circulation redistributes value rather than creating it. The second stage is money. Money is the #general_equivalent, the commodity in which the value of all other commodities is expressed. Once money exists, capital can be defined as value that expands itself through a circuit: money buys commodities, including labour power, which are used to produce new commodities that sell for more money than was originally advanced. The extra is #surplus_value, and it originates in production, not in the market. The third stage is #interest_bearing_capital. Once money can be lent, the lender advances money and receives more money back without personally organising any production at all. Marx writes this circuit as #M_M_prime, money begetting money. It is, he says, the most fetishised form of capital, because the intermediate stage where value is actually produced has disappeared from view. The lender sees only that money grows, as if growth were a natural property of money in the way that fruiting is a natural property of a pear tree. This appearance is what Marx meant by #commodity_fetishism applied to finance. The fourth stage is capitalisation. Once a regular income stream exists, and once claims on that income can be bought and sold, the market will price the claim by asking what sum of money, lent at the going rate of interest, would yield the same income. An annual income of one hundred, at a five percent rate of interest, will support a price of two thousand. Marx's insight is that this operation can be performed on any income stream whatsoever, whether or not any capital was ever advanced to produce it. A government bond is the classic case. The state borrowed money and spent it, often on wars, and the money is gone. Nothing productive stands behind the bond. Yet the bond has a price, it trades, it can be pledged as #collateral, and its owner receives an income. The capital, in Marx's phrase, is purely illusory, and yet the claim is entirely real. The fifth stage is fictitious_capital proper. Because the same underlying value can be represented several times over, once in the productive assets, once in the shares, once in the loan secured against the shares, once in the derivative written on the loan, the total volume of paper claims can grow far faster than the volume of value actually being produced. Marx describes this as capital seeming to double or treble itself. In an expansion, the growth of claims and the growth of production reinforce each other. In a contraction, the claims collapse back towards the value that actually exists, and the difference is destroyed. 3.2 The essential features From this sequence I extract five features which together constitute the Marxian category. These become the analytical criteria used later in the paper. Feature 1: Tradability of the claim. The asset must be a title that circulates independently of whatever it is a title to. Feature 2: Pricing by expectation. The price must be set by discounting expected future returns rather than by the cost of reproducing anything. Feature 3: Detachment from #real_accumulation. Movements in the price must be capable of diverging, systematically and for long periods, from movements in the productive base. Feature 4: Credit embedding. The asset must be usable as collateral, so that it becomes an input into further claim creation. This is what turns a speculative asset into a source of #systemic_risk. Feature 5: Destructibility without physical loss. In a crisis the paper value must be capable of being annihilated while the physical world remains intact. Notice what is not on the list. I have deliberately left out the requirement that the claim must be attached to a real productive asset, following Alves (2023), who argues that fictitious capital does not represent real capital. I have also, provisionally, left out the requirement of an actual revenue stream, and instead folded it into Feature 2 as pricing by expectation. Whether that move is legitimate is one of the central questions of the paper, and Section 11 returns to it. 3.3 Crisis in the Marxian frame Marx's theory of crisis, at least in its financial aspect, follows from the same sequence. During an expansion, credit allows purchases to run ahead of realised income. Paper claims are created against expected future income, and those claims are themselves used as security for more borrowing. The pyramid grows. It is sustained by the belief that the income will arrive. When that belief breaks, everyone tries to convert paper claims back into money at the same moment, and discovers that the money does not exist, because the claims were never more than promises. Prices collapse, collateral values fall, further sales are forced, and the collapse feeds itself. This is a #liquidation_spiral. At the end of it, a large quantity of paper wealth has been destroyed, which Marx describes as #devalorisation, and the productive apparatus has hardly been touched. The prediction that follows is precise and testable. In a financial crisis of this type we should observe: rapid growth of leverage during the boom; the recycling of assets as collateral; a trigger that undermines confidence rather than productive capacity; a run in which everyone attempts to exit simultaneously; forced liquidation that amplifies the fall; enormous destruction of recorded value; and comparatively little destruction of physical capacity. Sections 5 to 10 test this sequence against the crypto crashes of 2022. 4. Methodology 4.1 Research design This is a conceptual and interpretive study with an embedded comparative case component. It does not generate new price data or run new econometric tests. Its purpose is theoretical: to test the explanatory reach of an existing analytical category against a body of already published empirical evidence. This design is standard in #political_economy and in the philosophy of economics, and it is appropriate here because the question at issue is not what happened, which is well documented, but what kind of thing crypto assets are, which is a question of interpretation. 4.2 Data sources The evidence base consists of published, peer reviewed empirical research on crypto markets, together with the analytical reports of official financial institutions with access to supervisory data, principally the Bank for International Settlements, the Financial Stability Board and the International Monetary Fund. Priority is given to sources published between 2021 and 2026, so that the analysis reflects the market after the arrival of large scale DeFi, stablecoins and #institutional_adoption rather than the earlier and much simpler Bitcoin only market. This is #secondary_data analysis, and its reliability depends on the reliability of the underlying studies, which is discussed under limitations. 4.3 Analytical procedure The procedure has three steps. Step one is operationalisation. The five features of fictitious capital derived in Section 3.2 are turned into five yes or no questions that can be asked of any digital asset. Step two is application. Five crypto sub markets are examined in turn against the five criteria. The sub markets are selected to span the range of the asset class: a pure coin with no cash flow (#Bitcoin), a liability with a promised par value (stablecoins), a lending system that generates an actual yield (DeFi credit), a unique digital title (#NFTs), and a pure claim on price movement (#perpetual_futures). Step three is synthesis. The results are combined into a typology and used to build a stylised crisis sequence, which is then compared with the documented events of 2022. 4.4 Scope and boundaries The paper does not attempt to settle the truth of the labour theory of value, which is a separate argument requiring separate evidence. It assumes the theory for the purposes of the analysis and asks what follows if it is granted. Readers who reject the theory can still follow the argument about capitalisation and leverage, which does not depend on it, though they will reach a weaker version of the conclusion. The paper also excludes central bank digital currencies from the main analysis, because a #central_bank_digital_currency is a direct liability of the state and therefore belongs to the theory of credit money rather than the theory of fictitious capital, though a brief comment appears in Section 10. 5. Bitcoin: Capitalisation Without a Revenue Stream 5.1 What Bitcoin is, in economic terms Strip away the technology and #Bitcoin is a public ledger recording who controls which units of an artificially scarce entry. New units are issued to those who expend computing power to validate transactions, a process called #mining, and the issuance schedule is fixed in advance. A unit pays no interest, confers no ownership over any firm, and grants no claim over any income. Its holder can do exactly two things with it: transfer it to someone else, or hold it in the hope that someone else will later pay more for it. This makes Bitcoin an awkward case for the Marxian framework, and it is best to face the awkwardness directly rather than to smooth it over. 5.2 Applying the criteria Tradability. Clearly satisfied. Bitcoin exists only as a transferable title. Indeed it is nothing but a title, which is a stronger form of the criterion than Marx ever contemplated. Pricing by expectation. Satisfied, but in a degenerate way. The price is entirely a function of what buyers expect other buyers to pay. There is no revenue to discount, so the capitalisation formula has a numerator of zero. In effect the market performs the discounting operation on an expectation of #price_discovery itself. This is the crucial theoretical point of the whole paper and Section 11 develops it. Detachment from #real_accumulation. Overwhelmingly satisfied. The empirical literature finds that crypto prices are driven by speculative flows, sentiment and liquidity conditions rather than by any measure of underlying network usage or economic output, and that the assets show extreme volatility inconsistent with a stable store of value (Baur and Dimpfl, 2021). Makarov and Schoar (2022) document that real economy usage of the networks remains small relative to trading volumes, meaning that the vast majority of activity is people buying and selling the claim to each other. Credit embedding. Strongly satisfied, and increasingly so. Bitcoin functions as collateral throughout crypto lending, both centralised and decentralised, and since the approval of exchange traded products it has been absorbed into the collateral chains of regulated finance as well. Once an asset can be pledged, its price ceases to be a private matter for its holders and becomes a variable in other people's balance sheets. This is the transition from gambling to #systemic_risk. Destructibility without physical loss. Satisfied to an extreme degree. In 2022 the recorded value fell by more than seventy percent from its peak. The hardware, the electricity grid, the factories and the workforces of the world economy were unaffected. What vanished was a set of expectations that had been given a price. 5.3 The mining question A recurring argument holds that Bitcoin does have a labour value, because #proof_of_work mining consumes real electricity, real hardware and real human labour, and that its price therefore reflects a cost of production in the ordinary Marxian sense. The argument is tempting and, I think, wrong, for two reasons. First, in Marx's system the socially necessary labour time that regulates value is the labour required to reproduce a useful thing. Mining does not produce a useful thing; it produces a lottery ticket, and the difficulty of producing it is deliberately adjusted upwards whenever more people try. If the cost of production rises automatically to absorb whatever revenue is available, then cost is following price rather than governing it. The causal arrow runs backwards compared with a normal commodity. Second, the electricity and hardware consumed in mining are values transferred out of the productive economy, not values created within it. From a strict Marxian standpoint mining is a cost of #circulation, comparable to the labour of a security guard or a stockbroker's clerk: socially necessary given the way the system is organised, but not productive of new #surplus_value. The enormous #energy_consumption of the network is therefore not evidence that Bitcoin embodies value. It is evidence of how much value society is prepared to burn in order to maintain the fiction. 5.4 Interim verdict Bitcoin satisfies four of the five criteria fully and the fifth in a peculiar limiting form. It is a title priced by capitalisation of an expectation with no underlying revenue. If fictitious capital is capital whose value has become detached from production, then Bitcoin represents the terminal case of that detachment: not a weak link to production but no link at all. I will call this pure fictitious capital, and Section 11 argues that this is a coherent extension of Marx rather than a violation of him. 6. Stablecoins: Private Credit Money and the Anatomy of a Run 6.1 The promise of par A stablecoin is a token that promises to be worth one unit of an official currency, usually the United States dollar. There are two main designs. The collateralised design holds reserves of real assets, typically bank deposits and short term government securities, and promises to redeem tokens against those reserves. The algorithmic design holds no such reserves and attempts to hold the peg by offering an arbitrage: a token can always be swapped for one dollar worth of a second, freely floating token issued by the same system. In Marxian terms, both designs are attempts to create private credit money. The token is a liability of the issuer, and it circulates at par because holders believe it can be converted into money proper on demand. Marx wrote at length about exactly this structure in his discussion of bank notes and the credit system, and his conclusion was that private credit money circulates smoothly precisely as long as nobody tests it, and that its whole architecture becomes visible only at the moment when everybody tests it at once. 6.2 The Terra collapse as a Marxian crisis in miniature The collapse of the TerraUSD system in May 2022 is the cleanest natural experiment the crypto economy has produced. The system paid an unusually high and effectively subsidised yield on deposits of its stablecoin through an associated lending protocol, which attracted an enormous inflow of funds. The stablecoin's peg was maintained not by reserves but by a swap mechanism with a companion token whose price was itself supported by demand for the stablecoin. Briola and colleagues (2023) document the mechanics of the failure, and Liu, Makarov and Schoar (2023) analyse it as a run in the classical sense. The circularity is the point. The stablecoin was backed by the companion token; the companion token was valuable because the stablecoin was in demand; the stablecoin was in demand because it paid a yield; the yield was paid out of reserves that would eventually run out. Every element of the structure was a claim on another element of the structure. There was no external revenue anywhere in the loop. This is the pyramid Marx described, built in public, at speed, and with the ledger visible to anyone who cared to look. When confidence broke, the mechanism worked in reverse with brutal efficiency. Redemptions of the stablecoin minted new units of the companion token, which pushed its price down, which reduced the perceived backing of the stablecoin, which triggered more redemptions. Roughly forty billion dollars of recorded value evaporated within days. Badev and Watsky (2023) trace how the shock propagated through connected protocols. Later in the same year the failure of a major centralised exchange, itself heavily exposed to a token it had issued and was using as security against loans, produced a second wave of the same pattern, an episode examined in detail by Conlon, Corbet and Hu (2023). 6.3 Applying the criteria Stablecoins satisfy tradability, credit embedding and destructibility without hesitation. They are the settlement asset of the entire crypto economy, and they are pledged everywhere. They are ambiguous on pricing by expectation, because a fully reserved stablecoin is priced at par by construction, not by discounting. They are ambiguous on detachment, because a fully reserved issuer is holding real government securities, which are themselves claims on future taxation and therefore, in Marx's own analysis, fictitious capital of the most orthodox kind. That last observation deserves emphasis, because it is often missed. A fully reserved dollar stablecoin is a claim on a portfolio of treasury bills. A treasury bill is a claim on the future tax revenue of a state. So the safest instrument in the crypto economy turns out to be a second order title to fictitious capital: a claim on a claim on money that has not yet been collected from taxpayers who have not yet earned it. The distance from production has not been reduced by adding reserves. It has been formalised. 6.4 Reflexivity and the impossibility of a purely algorithmic peg The deeper lesson of the algorithmic design is that a system cannot bootstrap its own security out of its own liabilities. This is a Marxian point rather than a technical one. Value cannot be created in circulation. A closed loop of claims can redistribute value among its participants, and can go on doing so for as long as new participants arrive with money earned outside the loop, but it cannot generate value internally. Structures with this property display #Ponzi_dynamics whether or not anyone involved intended fraud, and the theory predicts that they must eventually fail, without being able to predict the date. 7. Decentralised Finance: Interest Bearing Capital Without Information 7.1 What DeFi actually does Decentralised finance replaces the bank with a piece of code. Lenders deposit tokens into a pool governed by #smart_contracts; borrowers take tokens out of the pool by depositing other tokens as security; an algorithm sets the interest rate according to how much of the pool is being used; and a liquidation mechanism sells the borrower's security automatically if its value falls too close to the size of the loan. At first sight this looks like the closest thing in the crypto economy to genuine interest bearing capital. Interest is actually paid. Someone borrows, someone lends, and a rate emerges. If Marx's circuit of money begetting money exists anywhere in crypto, it exists here. 7.2 But lending to whom, and for what? The question a Marxian analyst must ask about any interest payment is where the interest comes from. In the classical circuit, the industrial capitalist borrows money, buys machines and labour power, produces commodities, sells them at a profit and pays the lender a share of that profit. Interest is a division of surplus value. It has a source outside the financial system. DeFi lending has almost no such source. Borrowers in these markets are not building factories. Because the system cannot identify anyone, it cannot assess creditworthiness, and so it demands overcollateralisation: the borrower must pledge more value than they receive (Aramonte, Doerr, Huang and Schrimpf, 2022). This design has a decisive consequence. Nobody who lacks assets can borrow. The system is structurally incapable of financing an entrepreneur who has a good idea and no capital, which is precisely the function that credit performs in the productive economy. So who borrows, and why? Overwhelmingly, borrowers pledge crypto assets in order to obtain more crypto assets, so as to take a larger position in crypto assets. Borrowing is used to lever up a speculative bet, to arbitrage between venues, or to obtain liquidity without triggering a taxable sale. The interest paid is therefore not a share of surplus value extracted from production. It is a transfer from one speculator to another, funded out of expected trading gains, which in turn must come from some other trader's losses or from new money entering the system. DeFi thus reproduces the outward form of interest bearing capital while emptying it of its content. The productive middle term of the circuit is not merely hidden, as Marx said it was in ordinary finance. It is genuinely absent. 7.3 Procyclicality and the liquidation spiral Because loans are secured against volatile assets, borrowing capacity rises and falls with the price of those assets. When prices rise, the same holdings support larger loans, which are used to buy more of the same assets, which pushes prices up further. This is procyclicality in its purest form, and it operates without any of the dampening mechanisms that constrain bank lending. On the way down the mechanism reverses and accelerates. As collateral values fall, positions approach their liquidation thresholds. Liquidation is automatic and executed by code, without discretion, forbearance or negotiation. Automated sales push prices lower, which pushes further positions into liquidation. Aramonte, Huang and Schrimpf (2021) identify exactly this combination of high leverage, liquidity mismatch and interconnection, together with the absence of any institution capable of absorbing a shock, as the core vulnerability of the sector. The liquidation spiral is not merely present in DeFi. It has been written into the software as a design feature and given the name of a safety mechanism. 7.4 The decentralisation illusion and the governance question The claim that DeFi removes intermediaries does not survive contact with the evidence. Aramonte, Huang and Schrimpf (2021) show that governance is unavoidable, that #governance_tokens concentrate in a small number of hands, and that the protocols are therefore controlled by identifiable groups whose interests are not identical to those of their users. This is the #decentralisation_illusion. Ownership of the protocol confers the right to set fees and parameters, which is to say the right to levy a charge on everyone who uses the infrastructure. This is where the #techno_feudalism argument becomes useful. If the income of a protocol's owners consists of fees extracted from users of a network they control, then that income is closer to #monopoly_rent than to profit, and the governance token that entitles its holder to a share of those fees is closer to a title to land than to a share in a factory. The token is priced by capitalising the expected rent. That is a textbook instance of fictitious capital in Marx's sense, and, unlike Bitcoin, it even has a revenue stream. It is, ironically, the most orthodox fictitious capital in the entire crypto economy. 7.5 Applying the criteria DeFi lending positions and governance tokens satisfy all five criteria. Tokens trade; they are priced by capitalising expected fee income and expected #yield_farming returns; the fee income depends on speculative trading volume rather than on production; the tokens are pledged as security within the same system that issues them; and their value can be destroyed overnight, as it repeatedly has been, without a single physical asset being harmed. 8. Non Fungible Tokens: Digital Scarcity and the Aesthetics of Rent 8.1 The invention of scarcity An NFT is a ledger entry saying that a particular address controls a particular unique token, which usually points to a file stored somewhere else. The token does not confer copyright, does not confer possession of the file, and in most cases does not confer any income at all. What it confers is a socially recognised claim to be the owner of something that, technically, everyone can copy. The economic interest of NFTs is that they demonstrate, in laboratory conditions, that #digital_scarcity is entirely a social construction. There is no natural scarcity in a digital file. Scarcity had to be manufactured by convention and enforced by a ledger, in order that a price could exist at all. Marx's argument that value is a social relation which appears as a property of a thing has rarely had a more literal illustration. 8.2 Rent, not profit Where NFTs do generate income, it is typically through royalties paid to the original issuer on every subsequent resale. This is not profit derived from production. It is a toll levied on circulation. Varoufakis (2023) and Durand (2024) both argue that this form of income is becoming dominant in the digital economy, and the NFT is perhaps its purest specimen: a quasi legal right to charge a fee whenever a thing changes hands, attached to a thing that costs nothing to reproduce. The valuation of an NFT collection therefore rests on capitalising an expected stream of resale royalties, plus an expectation of price appreciation. When trading volume collapses, both components go to zero simultaneously, which explains the completeness of the market's collapse after 2022. 8.3 Manufactured volume The empirical literature on NFT markets documents extensive wash trading and insider dealing, in which parties transact with themselves to create the appearance of demand and to inflate recorded prices. Similar findings hold for crypto exchanges generally, where Cong and colleagues (2023) estimate that fabricated volume accounts for a very large share of reported activity on unregulated venues. The Marxian significance of this evidence is not simply that fraud occurred. It is that the price, which is the only thing an NFT has, is itself partly manufactured by the parties who benefit from it. In a market for a productive asset, manipulation distorts a price that would otherwise reflect something. In a market for a pure title with no revenue, manipulation is not a distortion of the signal. It is the signal. #Market_manipulation here is constitutive rather than parasitic. 8.4 Applying the criteria NFTs satisfy tradability, detachment, and destructibility completely. They satisfy pricing by expectation, where the expectation is of resale rather than of income. They satisfy credit embedding only partially, since illiquid and unique items make poor security, though lending against the best known collections did develop. The partial failure of the fourth criterion explains an important empirical fact: the NFT crash of 2022, though close to total in percentage terms, generated far less systemic contagion than the stablecoin and lending failures of the same year. Assets that are not embedded in credit chains can collapse without dragging anything else down with them. This is a testable implication of the framework, and the evidence confirms it. 9. Crypto Derivatives: Claims Upon Claims 9.1 The perpetual future The largest crypto market by volume is not the market in coins. It is the market in derivatives, and above all in perpetual futures, a contract with no expiry date that tracks the price of an underlying token and is settled continuously through a funding rate paid between long and short positions. These instruments are routinely traded with leverage of ten, twenty, fifty or even one hundred times the trader's deposit. A perpetual future on Bitcoin is a claim whose value derives from the price of an asset which is itself a claim on nothing. It is a second order fiction: a bet on the movement of a title that has no underlying revenue. Marx observed that in a developed credit system the same capital appears several times over in different hands and in different forms. Crypto derivatives extend the observation by allowing capital that never existed to appear several times over. 9.2 Leverage as an accelerator The function of leverage in this system is to permit a large exposure with a small deposit, and its consequence is that a small adverse price movement extinguishes the deposit entirely. When that happens the position is closed automatically, which means an order to sell arrives in the market at exactly the moment when prices are already falling. Because a very large number of positions are constructed similarly, the liquidations arrive in clusters, and each cluster pushes the price down to the level at which the next cluster triggers. This is the mechanism behind the characteristic shape of a crypto crash: not a gradual decline but a series of near vertical drops separated by pauses. The pattern is not evidence of irrationality. It is evidence of a leverage structure functioning exactly as designed. Official assessments identify precisely this dynamic, alongside opacity and interconnection, as the principal transmission channel for #contagion within the sector (Financial Stability Board, 2022; International Monetary Fund, 2022). 9.3 Applying the criteria Crypto derivatives satisfy every criterion in the strongest possible form. They are pure titles, priced entirely by expectation, wholly detached from production, deeply embedded in credit through margin and security chains, and capable of being annihilated instantly. If Bitcoin is fictitious capital in its pure form, the perpetual future is fictitious capital squared. 10. Synthesis: A Marxian Anatomy of the Crypto Crisis Cycle 10.1 The typology Table 1 summarises the results of Sections 5 to 9. Asset class Tradable title Priced by expectation Detached from production Embedded in credit Destructible without physical loss Revenue stream exists Bitcoin and similar coins Yes Yes, with no revenue term Completely Yes, increasingly Yes No Reserved stablecoins Yes Priced at par, not discounted Indirectly, through government debt Yes, systemically Yes, if reserves fail Yes, from reserves Algorithmic stablecoins Yes Yes, reflexively Completely Yes Yes, demonstrated No DeFi governance tokens Yes Yes, by capitalising fees Fees derive from speculation Yes Yes Yes, as rent NFTs Yes Yes, by expected resale Completely Partially Yes Sometimes, as royalty Perpetual futures Yes Entirely Completely Extremely Yes, instantly No Two conclusions follow from the table. The first is that the crypto economy is not a single object. It contains at least three distinct economic forms: pure speculative titles with no revenue, rent bearing titles with a real fee income, and private credit money. Analysing them together as one asset class, which most public commentary does, guarantees confusion. The second is that the criterion these assets fail most often is the existence of a revenue stream, and the criterion they satisfy most completely is detachment from production. This is the inverse of what Marx described for shares and bonds, which have a genuine revenue stream and only a partial detachment. The crypto economy has therefore not merely extended fictitious capital. It has inverted its internal composition. 10.2 The crisis sequence The Marxian model predicts a specific sequence, set out in Section 3.3. Table 2 compares the prediction with the documented events of 2022. Predicted stage Observed in the 2022 crypto crisis Credit expansion on the basis of expected future income Rapid growth of lending protocols and centralised lenders offering yields far above anything available in the productive economy Claims used as security for further claims Tokens pledged to borrow tokens to buy tokens, including tokens issued by the borrower's own affiliates Pyramiding of paper values far beyond any productive base Recorded market capitalisation approaching three trillion dollars against negligible cash flows Confidence shock rather than productive shock A depegging event and a loss of faith in reserves, not a fall in output anywhere Simultaneous attempt to convert claims into money Runs on stablecoins, lending platforms and exchanges within days Forced liquidation amplifying the fall Automated liquidations, margin calls and cascading sales Destruction of paper wealth Roughly two trillion dollars of recorded value erased Physical productive base unharmed No measurable loss of productive capacity attributable to the crash The correspondence is close enough to shift the burden of proof. A model written in the 1860s, applied to an asset class invented in 2009, reproduces the sequence of a crisis that occurred in 2022, stage by stage. Whatever one concludes about the labour theory of value, the theory of the credit cycle embedded in Marx's account is doing real explanatory work here. 10.3 Why the crisis stayed contained, and why that may not last The 2022 collapse did not become a general #financial_crisis. The Marxian framework explains why: the crypto system was not yet deeply embedded in the credit chains of the productive economy. Losses fell on holders of crypto claims, and the pyramid was largely self contained. The so called #crypto_winter that followed was severe within the sector and almost invisible outside it. The framework also identifies the condition under which this would change. The danger arises not from the size of crypto valuations but from the degree of their embedding in security chains that reach into regulated finance. As crypto assets are absorbed into exchange traded products, as banks take exposure, and as large issuers such as #Tether become significant holders of short term government securities, the situation reverses. A forced sale of reserves by a stressed stablecoin issuer would transmit stress directly into money markets. This mechanism is repeatedly flagged in official analysis (Financial Stability Board, 2023; Auer, Borio and others, 2025), and the Marxian framework converts it from a list of risks into a structural prediction: the more successfully crypto integrates with mainstream finance, the more certainly its next crisis will not be contained. 10.4 Policy implications Four #policy_implications follow, and none of them is technological. First, if the instability arises from the structure of leveraged claim creation rather than from the specific software, then rules aimed at the software will always arrive a step behind. Regulation should target leverage, the reuse of the same asset as security in several places at once, and the interface between crypto and regulated balance sheets. Second, since reported market capitalisation is calculated from prices formed in venues with substantial fabricated volume, official statistics that repeat these figures are reporting a number that is partly manufactured. Improving #transparency of trading data is a precondition for measuring the exposure at all. Third, #investor_protection cannot be reduced to disclosure. In a market where the asset has no cash flow, there is nothing meaningful to disclose about fundamentals, and disclosure of risk factors has repeatedly failed to prevent losses among #retail_investors. The relevant protections are structural: limits on the leverage offered to households, and separation of custody from proprietary trading. Fourth, a central bank digital currency would not eliminate the phenomena described here, because the object of speculation is not the payment technology but the capitalisation of expectations. It would, however, remove one of the principal justifications offered for private stablecoins. 11. Discussion: Extending the Category 11.1 The problem restated The analysis has produced a paradox. Crypto assets behave like fictitious capital, crash like fictitious capital, and are embedded in credit like fictitious capital, but most of them lack the one thing Marx's definition seems to require, which is a revenue stream to capitalise. If we insist on the definition, we must say that Bitcoin is not fictitious capital, and then we are left without an explanation for why it behaves exactly as if it were. If we abandon the definition, we risk emptying the category of content and turning it into a synonym for anything speculative. There is a third path, and it is the theoretical proposal of this paper. 11.2 Separating the substance from the mechanism Marx's fictitious capital has two components that are usually run together. The first is the substance: an expected future income, arising somewhere, which is what the claim is a claim to. The second is the mechanism: a set of social operations by which an expectation is converted into a price, the price into a tradable title, the title into security for borrowing, and the borrowing into further purchases of the title. Call this the machinery of capitalisation. Marx's own examples show that the substance is far less essential than it looks. In the case of public debt, the money borrowed has been spent and destroyed, and no capital exists behind the bond at all. Marx says so explicitly, calling the capital purely illusory. Alves (2023) develops precisely this reading and concludes that fictitious capital need not represent real capital. What Marx did not consider, because it did not exist, is the possibility that the mechanism could operate on an expectation with no income term at all. That is what crypto does. The market takes an expectation, prices it, makes the price tradable, lends against the title, and uses the borrowed funds to buy more of the title. Every step of the machinery of capitalisation is executed faithfully. Only the object at the centre is missing. 11.3 A proposed extension: fictitious capital in its pure form I therefore propose the following extension. Fictitious capital should be defined by its mechanism rather than by its object. On this definition, a title is fictitious capital when its price is produced by the capitalisation machinery, when it circulates as an independent claim, and when it functions as security for the creation of further claims. The presence or absence of an underlying income is a variable that determines the degree of purity, not a condition of membership. This yields a spectrum rather than a binary. At one end sit corporate shares: a real revenue, a real productive asset, and a capitalised price that can nevertheless detach from both for long periods. In the middle sit government bonds: a real revenue, no productive asset, a capitalised price. Further along sit DeFi governance tokens: a real fee income, no productive asset, and an income that itself derives from speculation. At the far end sit Bitcoin and the perpetual future: no income, no asset, and nothing left but the machinery. Crypto assets are therefore not an exception to Marx's category. They are its limit case, the point at which the tendency he identified reaches completion. He wrote that in interest bearing capital the source of value becomes invisible, so that money appears to breed money by its own nature. In crypto, the source has not merely become invisible. It has been abolished, and money appears to breed money with nothing whatsoever in between. The fetish has become the entire object. 11.4 What this explains that other theories do not Three empirical features of crypto markets are difficult to explain from a conventional finance standpoint but follow naturally from this extension. The first is the persistence of extreme volatility. Standard asset pricing expects volatility to fall as a market matures, as information improves and as institutional participants arrive. In crypto it has not fallen anything like as much as maturity would predict, and the reason is structural rather than informational: when there is no income term, there is nothing to anchor the price, so the only anchor available is the current opinion of other holders. Baur and Dimpfl (2021) show that this volatility is severe enough to undermine the asset's own claimed monetary functions. The second is the recurrence and the similarity of the crashes. Each crypto cycle produces a different proximate cause, and each produces the same shape. A theory that treats each crash as a governance failure must explain the repetition as coincidence. A theory that locates the cause in the structure of leveraged claim creation predicts the repetition directly. The third is the centrality of manipulation. If a large share of the volume in these markets is fabricated, as the empirical work indicates, then in an income bearing market this would be a scandal that distorts the signal. In a market with no income, the manufactured price is not a distortion of the fundamental. It is the only fundamental there is. The extension explains why manipulation is not peripheral to crypto but load bearing. 11.5 The counterargument, and a response The strongest objection to this argument is that it is unfalsifiable. If crypto has no income and is fictitious capital, and shares have income and are also fictitious capital, what would not be fictitious capital? The answer is that the framework has five criteria and they do real work. A physical machine owned outright by the firm that uses it, and not pledged to anyone, fails four of them. Gold held in a vault fails the credit embedding criterion when it is not pledged, and satisfies it when it is, which is exactly the right result. An NFT satisfies four criteria and only partially satisfies the fifth, and the framework predicted, correctly, that this would make the NFT crash locally severe and systemically harmless. The category is discriminating. A second objection is that the analysis proves too little, because a follower of Keynes or Minsky could say much the same thing without any of the Marxian machinery. This is fair, and worth conceding openly. Minsky's account of financial fragility, in which stability breeds leverage and leverage breeds collapse, describes the same sequence. The difference is in what the two theories say about the source of the money that flows into the pyramid. Minsky explains the dynamics of the pyramid. Marx explains why the pyramid must, in the end, be settled against value produced elsewhere, and therefore why no amount of internal financial engineering can make a closed loop of claims self sustaining. That is a stronger claim, and it is the claim that the Terra episode tested and confirmed. 11.6 The rent question and the techno feudalism debate A final theoretical issue deserves attention. If the income of the crypto sector consists mainly of fees and tolls levied on users of privately controlled infrastructure, then it may be that #rent rather than profit is the appropriate category, and that the relevant Marxian text is the theory of ground rent rather than the theory of fictitious capital. Varoufakis (2023) and Durand (2024) push this argument to the conclusion that the system has ceased to be capitalism at all. The evidence assembled here does not support so strong a conclusion. Rent and fictitious capital are not alternatives; they are complements. Ground rent is capitalised into a land price, and the resulting title is itself fictitious capital. The same is true of a governance token. What crypto shows is not the replacement of capital by rent but the growth of a layer of titles whose income, where it exists at all, is rent rather than profit, and whose prices are set by the same capitalisation machinery that prices everything else. That is an intensification of the tendency Marx described, not a departure from it. 12. Limitations and Directions for Future Research 12.1 Limitations Four limitations should be stated plainly. First, this is a theoretical study resting on #secondary_data. It has not generated new measurements, and its conclusions are only as reliable as the studies it draws on. Where those studies rest on exchange reported data, the wash trading literature gives grounds for caution about the underlying numbers, and this caution applies to the present paper as much as to any other. Second, the analysis assumes the labour theory of value rather than defending it. Readers who reject that theory can accept the argument about capitalisation, leverage and crisis, which does not depend on it, but they will not accept the argument about the absence of value creation in circulation, which does. Third, the crypto market is a moving target. The rise of tokenised government securities, the arrival of exchange traded products and the integration of stablecoins into payment systems are all changing the object of study as the study is written. Some of the conclusions in Section 10 may date quickly, and the framework should be reapplied as the market changes. Fourth, the paper treats crypto largely as a financial phenomenon and says relatively little about the labour that builds and maintains the infrastructure, about the environmental costs of #proof_of_work, or about the distributional consequences of the transfers documented here. Each of these deserves a separate study. 12.2 Directions for future research Five lines of work follow naturally. The first is empirical measurement of the ratio between recorded crypto capitalisation and any measure of underlying cash flow, tracked over the cycle, which would give a quantitative index of the degree of fictitiousness and allow the theory to be tested against price data. The second is a study of security reuse, tracing the same asset as it is pledged in several places at once, which would put a number on the pyramiding that the theory predicts. The third is comparative history: the framework developed here should be applied to the railway manias of the nineteenth century, to the securitised mortgage products of the 2000s, and to crypto, in order to test whether the same crisis sequence appears in all three. The fourth is the study of the growing links between crypto and regulated finance, which the framework identifies as the decisive variable determining whether the next crypto crisis stays contained. The fifth is a study of who actually gains and who actually loses in these markets. If the interest and the trading profits are transfers rather than shares of newly produced surplus value, then the sector is redistributive by construction, and the direction of the redistribution is an empirical question of real social importance. 13. Conclusion This paper set out to discover whether a nineteenth century category could explain a twenty first century market. The answer is that it can, but only after being modified in one specific respect. Marx's account of fictitious_capital identifies a machinery: an expectation is capitalised into a price, the price becomes a tradable title, the title becomes security for borrowing, and the borrowing buys more titles. He argued that this machinery allows paper claims to multiply far beyond the value actually being produced, that the resulting structure is inherently unstable, and that its collapse destroys enormous quantities of recorded wealth while leaving the physical productive base untouched. Every stage of that machinery is present in the crypto economy, and the crisis of 2022 followed the predicted sequence with a fidelity that is difficult to attribute to chance. Leverage expanded against expected future income; claims were pledged to create further claims; a confidence shock, not a productive shock, triggered a run; automated liquidation amplified the fall; roughly two trillion dollars of recorded value was destroyed; and not one factory closed as a result. What crypto adds to Marx is not a refutation but a limiting case. In shares, the fiction is that a paper title is itself productive. In government bonds, the fiction is deeper, because the capital behind the title was spent long ago. In crypto, the fiction is complete, because there was never any income to be a claim to. The machinery of capitalisation now runs on an expectation of nothing but the expectation itself. This is why the volatility is so extreme, why the crashes recur, and why #market_manipulation is not an aberration in these markets but part of their basic operation. For students of #economics and #finance, the practical lesson is uncomfortable but useful. The crypto economy is not a strange island governed by its own laws. It is an unusually clear window onto the ordinary workings of the credit system, in which the usual layers of productive activity have been stripped away and the machinery is left running in the open, visible, fast, and public. Anyone who wants to understand how #speculation, #collateral and leverage combine to produce a #bubble and then a #crisis_theory that explains it will find no better teaching material. That, in the end, may be the most durable contribution of the digital asset economy: not a new form of money, but an exceptionally honest demonstration of how the old one actually works. References Alves, C. (2023). Fictitious capital, the credit system, and the particular case of government bonds in Marx. New Political Economy, 28(3), 398 to 415. Aramonte, S., Doerr, S., Huang, W., and Schrimpf, A. (2022). DeFi lending: intermediation without information? BIS Bulletin No. 57. Bank for International Settlements, Basel. Aramonte, S., Huang, W., and Schrimpf, A. (2021). DeFi risks and the decentralisation illusion. BIS Quarterly Review, December 2021, 21 to 36. Auer, R., Borio, C., Hofmann, B., and Shin, H. S. (2025). Cryptocurrencies and decentralised finance: functions and financial stability implications. BIS Papers No. 156. Bank for International Settlements, Basel. Auer, R., Haslhofer, B., Kitzler, S., Saggese, P., and Victor, F. (2024). The technology of decentralized finance (DeFi). Digital Finance, 6(1), 55 to 95. Badev, A., and Watsky, C. (2023). Interconnected DeFi: ripple effects from the Terra collapse. FEDS Notes. Board of Governors of the Federal Reserve System, Washington DC. Bank for International Settlements (2022). The future monetary system. In BIS Annual Economic Report 2022, Chapter III. Bank for International Settlements, Basel. Baur, D. G., and Dimpfl, T. (2021). The volatility of Bitcoin and its role as a medium of exchange and a store of value. Empirical Economics, 61(5), 2663 to 2683. Briola, A., Vidal-Tomas, D., Wang, Y., and Aste, T. (2023). Anatomy of a stablecoin's failure: the Terra-Luna case. Finance Research Letters, 51, 103358. Comyn, C. (2026). Fictitious money: cryptocurrency as a social form. New Political Economy, advance online publication. Cong, L. W., Li, X., Tang, K., and Yang, Y. (2023). Crypto wash trading. Management Science, 69(11), 6427 to 6454. Cong, L. W., Li, Y., and Wang, N. (2021). Tokenomics: dynamic adoption and valuation. Review of Financial Studies, 34(3), 1105 to 1155. Conlon, T., Corbet, S., and Hu, Y. (2023). The collapse of FTX: the end of cryptocurrency's age of innocence. British Accounting Review, 56(1), 101277. Durand, C. (2024). How Silicon Valley Unleashed Techno-feudalism. Verso, London. Fang, F., Ventre, C., Basios, M., Kanthan, L., Martinez-Rego, D., Wu, F., and Li, L. (2022). Cryptocurrency trading: a comprehensive survey. Financial Innovation, 8(1), 13. Financial Stability Board (2022). Assessment of Risks to Financial Stability from Crypto-assets. Financial Stability Board, Basel. Financial Stability Board (2023). The Financial Stability Risks of Decentralised Finance. Financial Stability Board, Basel. Fuchs, C. (2022). Digital Capitalism: Media, Communication and Society, Volume Three. Routledge, London. Harvey, D. (2023). A Companion to Marx's Grundrisse. Verso, London. International Monetary Fund (2022). The rapid growth of fintech: vulnerabilities and challenges for financial stability. In Global Financial Stability Report, April 2022, Chapter 3. International Monetary Fund, Washington DC. Lapavitsas, C. and the EReNSEP Writing Collective (2023). The State of Capitalism: Economy, Society, and Hegemony. Verso, London. Liu, J., Makarov, I., and Schoar, A. (2023). Anatomy of a run: the Terra Luna crash. NBER Working Paper No. 31160. National Bureau of Economic Research, Cambridge MA. Makarov, I., and Schoar, A. (2022). Cryptocurrencies and decentralized finance (DeFi). Brookings Papers on Economic Activity, Spring 2022, 141 to 215. Marx, K. (1991 [1894]). Capital: A Critique of Political Economy, Volume III. Penguin Classics, London. Varoufakis, Y. (2023). Technofeudalism: What Killed Capitalism. The Bodley Head, London. #MarxianEconomics #FictitiousCapital #CryptoEconomics

Commercial organisations still decide most projects on a narrow set of financial tests: net present value, internal rate of return, payback period, and a short risk register. These tests are useful, but they were designed for a world in which the costs a firm imposed on society, and the value it created outside its own accounts, could be treated as somebody else's problem. That world no longer exists. Regulators now demand sustainability disclosure, investors ask for evidence of real world outcomes, employees ask what a project is for, and communities can delay or stop a project that ignores them. This article asks a practical question: how can societal impact be brought inside the #business_case rather than bolted on afterwards? Using an integrative review of literature published mainly between 2020 and 2025, drawn from strategic management, accounting, finance, project studies, and impact measurement, the article does three things. First, it explains why the traditional business case fails, and shows that the failure is structural rather than a matter of goodwill. Second, it maps the main families of impact metrics now in use, from ESG ratings to impact weighted accounts, and sets out honestly what each can and cannot do. Third, it proposes the Integrated Project Value framework, a five stage process that moves from scoping and materiality, through baseline and theory of change, to valuation, decision rules, and monitoring with a learning loop. The framework does not require every impact to be converted into money. It requires that impacts be identified, measured against a credible counterfactual, weighted transparently, and carried into the decision in a way that can be audited later. The article discusses barriers, including data quality, comparability, incentive design, and the risk of impact washing, and closes with a research agenda. The contribution is a decision ready structure that treats #societal_impact as evidence for judgement rather than as decoration for a report. Keywords: business case; project appraisal; societal impact; impact measurement; ESG; stakeholder theory; sustainable project management; corporate purpose; decision making; double materiality 1. Introduction Every organisation that spends money on projects has some way of deciding which projects deserve the money. In most commercial firms that way is the business case. A sponsor estimates the cash the project will bring in, estimates the cash it will consume, discounts the difference back to today, compares the answer with a hurdle rate, and either wins approval or does not. The logic is clean, teachable, and defensible in front of a board. The logic is also incomplete, and the gap has become impossible to ignore. A factory that meets its return target can still exhaust an aquifer. A cost saving in a supply chain can rest on labour conditions the buying firm would not accept in its own offices. A software platform can hit every commercial milestone while eroding the mental health of the people who use it. None of these effects appear in the cash flow model, because the model was never designed to see them. Economists have a name for this blind spot, #externalities, and for most of the twentieth century the standard answer was that governments would correct them through taxes, fines, and rules, leaving firms free to maximise #shareholder_value. That division of labour is breaking down for three reasons. The first is regulatory. Sustainability reporting has moved from voluntary to mandatory in many jurisdictions, and reporting requirements have a way of travelling upstream into decision making, because a firm cannot report on what it never decided to measure. The second is financial. Investors increasingly price climate risk, transition risk, litigation risk, and reputational risk, which means that societal effects that used to sit outside the model now leak into the cost of capital. The third is social. Communities, employees, and customers have more information and more coordination power than they had a generation ago, and they can impose real delays and real costs on a project that they consider illegitimate. In project terms, #legitimacy has become a schedule risk. The response of many firms has been to build a parallel process. The finance team produces the business case, and a sustainability team produces a sustainability appraisal, and the two documents meet, if at all, at the last governance gate. This arrangement satisfies nobody. It gives the impact analysis no leverage over the decision, because by the time it appears the decision has effectively been made. It also gives the financial analysis a false sense of completeness, because the risks that impact analysis would have surfaced remain outside the numbers until they arrive as surprises. This article argues for integration rather than parallelism. It proposes that societal impact be treated as an ordinary input to #project_evaluation, subject to the same demands for evidence, the same scepticism, and the same audit trail as revenue forecasts. That is a stronger claim than it may appear, because it forbids two comfortable positions. It forbids the position that impact is too soft to measure, which is usually a way of avoiding the effort. It also forbids the position that impact can be reduced to a single score that solves the decision automatically, which is usually a way of avoiding the judgement. The research questions are as follows. RQ1. Why does the conventional business case systematically exclude societal effects, and what are the consequences of that exclusion for commercial decision quality? RQ2. What families of societal impact metrics exist, and what are the evidenced strengths and weaknesses of each for use inside a project appraisal? RQ3. What process would allow impact evidence to enter a commercial project decision in a way that is credible, proportionate, and auditable? RQ4. What barriers stand in the way, and what conditions make integration more likely to succeed? The article proceeds as follows. Section 2 sets out the background. Section 3 reviews the literature across five streams. Section 4 develops the theoretical framing. Section 5 explains the method. Section 6 presents the analysis of metric families, barriers, and enablers. Section 7 sets out the proposed framework in detail. Section 8 illustrates the framework through three stylised project types. Section 9 discusses implications. Section 10 states limitations, Section 11 proposes a research agenda, and Section 12 concludes. 2. Background: How the Business Case Became Narrow It is worth remembering that the narrowness of the business case is a historical achievement rather than a natural fact. Discounted cash flow analysis solved a genuine problem. Before it, capital allocation inside large firms was often a contest of seniority and persuasion. Converting proposals into a common unit, money over time, allowed comparison across very different opportunities and imposed a discipline that improved decisions. The trouble is that the common unit does its work by excluding whatever cannot be priced. A cash flow model is not neutral about what matters. It is a claim that what matters is the cash accruing to the firm. Everything else is either a constraint to be satisfied at minimum cost, such as compliance, or an intangible to be mentioned in the narrative section that nobody reads. Two further habits reinforced the narrowness. The first is the treatment of risk. Conventional appraisal handles uncertainty by adjusting the discount rate or by running sensitivity analysis on a few variables, usually price and volume. Societal risks tend not to be in that list, so a project can pass a sensitivity test while remaining exposed to a community objection that stops it for a year. The second is the treatment of time. Discounting at commercial rates makes effects that arrive decades later almost invisible, which is precisely when many environmental and health effects arrive. Alongside these habits, a large body of practice grew up under the label of corporate social responsibility. Much of it was sincere. Much of it was also disconnected from the capital allocation process, funded from a separate budget, reported in a separate document, and evaluated, if at all, by counting activities rather than outcomes. Barnett, Henriques, and Husted argue that this design flaw explains why so many well intentioned corporate social initiatives produce so little measurable social change, and they call for initiatives to be designed backwards from the intended social outcome rather than forwards from the available budget (Barnett, Henriques, and Husted, 2020). The rise of #ESG in investment practice looked, for a while, like the correction. Capital would reward firms that scored well on environmental, social, and governance factors, and firms would respond by improving. Something like that has happened, but the mechanism has proved weaker and messier than its advocates hoped, for reasons the literature has since documented in detail and which Section 3 discusses. Meanwhile, in public sector appraisal, a parallel tradition had been quietly doing what commercial appraisal did not. Cost benefit analysis in government has long attempted to value time saved, lives saved, noise avoided, and habitats lost, using shadow prices and standardised valuation guidance. The United Kingdom Green Book, for example, requires appraisal of social value across the whole of society rather than the sponsoring department alone, and requires that unmonetisable effects still be presented to decision makers rather than dropped (HM Treasury, 2022). The Dasgupta Review made a related argument for natural capital, insisting that economic reasoning which treats nature as free is not rigorous economics but bad accounting (Dasgupta, 2021). The commercial sector has been slower to adopt these ideas, partly because it lacks the mandate, and partly because it lacks the shared valuation infrastructure that public appraisal has built over decades. The result is an odd asymmetry. A government deciding whether to build a road is expected to estimate the value of noise, air quality, and travel time for people who will never pay a toll. A corporation deciding whether to build a plant of similar size is expected to estimate only its own cash flows. The argument of this article is that the asymmetry is no longer sustainable, and that #accountability for societal effects will arrive in the commercial business case one way or another, either by design or by crisis. 3. Literature Review The relevant literature sits in five streams that rarely speak to each other. This section reviews each and then identifies the gap that the article addresses. 3.1 Purpose, stakeholders, and the theory of the firm The intellectual case for widening the business case rests on #stakeholder_theory, which holds that a firm creates value through its relationships with employees, customers, suppliers, communities, and financiers, and that treating any of these as a mere means is both ethically weak and commercially fragile. Freeman, Phillips, and Sisodia address the standard objection, that stakeholder theory gives managers no way to make trade offs, by arguing that the tensions are real and productive rather than fatal, and that the job of management is precisely to find directions in which stakeholder interests move together (Freeman, Phillips, and Sisodia, 2020). Edmans makes the economic version of the argument. In his account, a firm that invests in its workforce, its customers, and its environment is not sacrificing profit but growing the pie from which profit is cut, and the evidence he assembles suggests that such firms have historically outperformed rather than underperformed (Edmans, 2020). Mayer takes the argument further into corporate law and governance, proposing that the purpose of a corporation is to produce profitable solutions to the problems of people and planet, and not to profit from producing problems (Mayer, 2021). George, Haas, McGahan, Schillebeeckx, and Tracey review the fast growing #corporate_purpose literature and note that the field has produced more advocacy than evidence, and that purpose remains loosely defined and inconsistently measured (George, Haas, McGahan, Schillebeeckx, and Tracey, 2023). The counter position deserves respect rather than dismissal. Bebchuk and Tallarita argue that stakeholder governance, as currently practised, is largely illusory, because managers retain discretion, face no enforceable duty to any stakeholder, and can use stakeholder language to insulate themselves from accountability without changing behaviour (Bebchuk and Tallarita, 2020). Their argument matters for this article, because it identifies exactly the failure mode that an integrated business case must avoid. If impact language enters the appraisal without measurement, without weighting, and without an audit trail, it will function as rhetoric, and rhetoric is worse than silence because it is harder to challenge. Hart and Zingales offer a middle route, arguing that firms should maximise shareholder welfare rather than shareholder market value, on the ground that shareholders are also citizens with preferences over social outcomes, and that where the firm can address a harm more cheaply than individuals can, it should (Hart and Zingales, 2022). This provides a defensible foundation for including impact in the business case without abandoning the idea that owners have a claim on the decision. 3.2 Measurement in the ESG tradition If impact is to enter the appraisal, it must be measured, and the most visible measurement infrastructure is the ESG ratings industry. The evidence here is sobering. Berg, Kolbel, and Rigobon show that ratings from major providers diverge substantially, and decompose the divergence into differences in scope, measurement, and weighting, with measurement differences accounting for the largest share (Berg, Kolbel, and Rigobon, 2022). Christensen, Serafeim, and Sikochi find that greater disclosure, contrary to intuition, is associated with greater rather than lesser disagreement between raters, which suggests that more data alone will not resolve the problem (Christensen, Serafeim, and Sikochi, 2022). #rating_divergence is not a temporary defect to be fixed by better software. It reflects genuine disagreement about what should count and how much. The consequences for practice are direct. If a firm cannot rely on an external score to tell it whether a project is good for society, it cannot outsource the judgement, and it must build enough internal capability to make the judgement itself. Edmans presses the point in a more provocative form, arguing that ESG has been oversold as a distinct discipline, that it is neither necessary nor sufficient for long term value, and that treating it as a separate silo distorts decisions and invites box ticking (Edmans, 2023). Gillan, Koch, and Starks, reviewing the corporate finance evidence, find real but conditional relationships between ESG practices and firm outcomes, with effects depending heavily on context, measurement, and the mechanism assumed (Gillan, Koch, and Starks, 2021). The investment channel is also weaker than commonly assumed. Kolbel, Heeb, Paetzold, and Busch distinguish between changing a company's cost of capital, engaging with management, and creating new markets, and conclude that the evidence for investor impact is strongest for engagement and weakest for the simple act of screening portfolios (Kolbel, Heeb, Paetzold, and Busch, 2020). Heeb, Kolbel, Paetzold, and Zeisberger add an uncomfortable experimental finding, that investors are willing to pay for impact but that their willingness does not increase much with the amount of impact delivered, which implies that a warm feeling rather than an outcome is being purchased (Heeb, Kolbel, Paetzold, and Zeisberger, 2023). Pastor, Stambaugh, and Taylor and, separately, Pedersen, Fitzgibbons, and Pomorski model the equilibrium consequences, showing that when enough investors have sustainability preferences, green assets can command lower expected returns precisely because they are valued for non financial reasons (Pastor, Stambaugh, and Taylor, 2021; Pedersen, Fitzgibbons, and Pomorski, 2021). For a project sponsor, the practical reading of this body of work is blunt. The capital market will not reliably reward a good project for being good. Some of the value must be captured through operational channels, such as risk reduction, licence to operate, talent retention, and customer loyalty, and some must be justified on grounds other than return. Pretending otherwise sets the business case up to fail its own test. 3.3 Reporting, materiality, and the risk of decoration The reporting literature explains why measurement so often fails to change behaviour. Pucker argues that the growth of sustainability reporting has outpaced the growth of sustainability, and that voluminous disclosure can substitute for action rather than driving it (Pucker, 2021). Adams and Abhayawansa examine the calls to harmonise the reporting landscape and warn that harmonisation aimed only at investors risks narrowing what is reported to what affects enterprise value, losing the effects on society that motivated the reporting in the first place (Adams and Abhayawansa, 2022). This concern is what #double_materiality is meant to address. Baumuller and Sopp trace the shift in European regulation from non financial reporting to sustainability reporting and set out the double materiality principle, under which a topic matters if it affects the firm financially, or if the firm affects society and the environment through it, or both (Baumuller and Sopp, 2022). Double materiality is more than a reporting concept. Carried into appraisal, it becomes a scoping rule with teeth, because it forces the project team to identify effects on society even where those effects do not, at least yet, threaten the firm. The literature on #greenwashing supplies the discipline needed to keep such claims honest. De Freitas Netto and colleagues, reviewing the concept, distinguish claim greenwashing, which misleads through words, from executional greenwashing, which misleads through imagery and association, and note the difficulty of detection when verification is weak (de Freitas Netto, Sobral, Ribeiro, and Luz Soares, 2020). Applied to project appraisal, the lesson is that any impact claim entering the business case should carry a source, a method, and an assurance status, exactly as a revenue forecast carries assumptions that an auditor can test. 3.4 The impact measurement tradition A separate stream, rooted in social enterprise, philanthropy, and development, has spent decades on the problem of measuring social change. Its core tools are the theory of change, which states how activities are expected to produce outcomes; the logic model, which structures inputs, outputs, outcomes, and impacts; and the counterfactual, which asks what would have happened anyway. Corvo, Pastore, Manti, and Iannaci map the models in use and find a crowded field with limited comparability, in which method choice is often driven by the reporting needs of funders rather than by the learning needs of the organisation (Corvo, Pastore, Manti, and Iannaci, 2021). Two ideas from this tradition are indispensable for commercial appraisal. The first is additionality. An impact only counts if the project caused it. A firm that hires local staff in a full employment region has moved employment rather than created it. Without a counterfactual, #impact_measurement collapses into attribution of everything good that happens nearby. The second is the distinction between outputs and outcomes. Training delivered is an output. Skill acquired and retained and used is an outcome. Reporting outputs while claiming outcomes is one of the commonest failures in practice. Busch and colleagues, writing on impact investing, argue for a reorientation of the field around investor contribution and verified outcomes rather than around intention, and note that the label impact has spread far faster than the practice of measuring it (Busch, Bruce-Clark, Derwall, Eccles, Hebb, Hoepner, and colleagues, 2021). Their diagnosis applies with equal force to corporate projects that describe themselves as impactful. The most ambitious attempt to bridge impact measurement and financial accounting is the #impact_weighted_accounts programme. Serafeim and Trinh propose a framework for valuing the impact of products on customers and society in monetary units so that the results can be placed alongside financial statements (Serafeim and Trinh, 2020). Freiberg, Park, Serafeim, and Zochowski extend the logic to environmental impact, showing that firm level environmental damage can be monetised at scale using available data, and that doing so changes the ranking of firms within industries (Freiberg, Park, Serafeim, and Zochowski, 2021). Serafeim's broader argument is that purpose and profit are compatible but that compatibility must be demonstrated with evidence rather than asserted (Serafeim, 2022). Kaplan and Ramanna offer a complementary technique for one domain, arguing that greenhouse gas emissions should be accounted for like a liability that is passed along the value chain, using audited transfers rather than estimates and averages (Kaplan and Ramanna, 2021). Monetisation of this kind is powerful and dangerous in equal measure. It is powerful because it lets impact enter the same arithmetic as cash. It is dangerous because it hides contested value judgements inside apparently technical coefficients, and because a single number invites the reader to stop thinking. 3.5 Project studies and the meaning of success The fifth stream is the one that matters most for implementation, because projects are where strategy becomes spending. Traditional project success was defined by the iron triangle of time, cost, and scope. That definition has been under sustained attack for years, and the critique has now matured. Ika and Pinto argue for a re meaning of #project_success, extending it beyond delivery efficiency to include the value produced for a range of stakeholders over time, and note that a project can be efficiently delivered and still be a failure in every sense that matters (Ika and Pinto, 2022). Zwikael and Meredith separate the success of the project from the performance of the people who led it, which is important because it removes an incentive to define success as whatever the manager controlled (Zwikael and Meredith, 2021). Locatelli and colleagues, in a manifesto for project management research, call for the field to engage with grand challenges and to move beyond the concerns of the delivery organisation towards the consequences of projects for society (Locatelli, Ika, Drouin, Muller, Huemann, Soderlund, Geraldi, and Clegg, 2023). Ika, Soderlund, Munro, and Landoni show what commercial project management can learn from international development, where the question of whether a project actually helped anybody has been the central question for decades (Ika, Soderlund, Munro, and Landoni, 2020). Work specifically on sustainable project management identifies a persistent implementation gap. Silvius and Schipper find that project managers address sustainability issues when they perceive personal responsibility and when the organisational context supports it, and that formal procedures alone are weak motivators (Silvius and Schipper, 2020). Sabini and Alderman describe the resulting position as paradoxical, with project professionals expected to deliver sustainability outcomes while being evaluated on cost and schedule, and conclude that the paradox cannot be resolved by exhortation alone (Sabini and Alderman, 2021). Two further strands of project research bear directly on the argument. Flyvbjerg documents the behavioural biases that distort appraisal, notably optimism bias and strategic misrepresentation, and shows that these biases inflate benefits and deflate costs in systematic ways (Flyvbjerg, 2021). Flyvbjerg and Gardner distil the practical implications, arguing for slow deliberate planning, reference class forecasting against actual outcomes of similar projects, and modular delivery that allows learning (Flyvbjerg and Gardner, 2023). These findings are a warning for anyone adding a new benefit category to a business case. If financial benefits are already overstated by optimism and by incentives to win approval, then adding a softer and less verifiable benefit category creates a new and attractive place to hide. 3.6 The research gap Each stream contributes something and each is incomplete. The purpose literature says why the business case should widen but not how. The ESG literature supplies data infrastructure whose reliability it has itself called into question. The reporting literature explains what must be disclosed after the fact but says little about appraisal before the fact. The impact measurement tradition has the right methods but was built for organisations whose primary goal is social, not for firms that must reconcile impact with return. Project studies has begun to redefine success but has not yet specified the appraisal mechanics that would make the redefinition operational. The gap is therefore procedural rather than conceptual. What is missing is a #conceptual_framework that takes the evidence standards of impact measurement, the materiality logic of sustainability reporting, the scepticism of the ESG critics, and the behavioural realism of project research, and combines them into a sequence a project sponsor can actually follow inside a stage gate process. That is the contribution attempted here. 4. Theoretical Framing The framework proposed in Section 7 rests on four theoretical commitments, stated here so that they can be criticised. Value is created in relationships, not extracted from them. Following stakeholder theory, the article treats the firm as a set of relationships whose health determines its capacity to create value over time (Freeman, Phillips, and Sisodia, 2020; Edmans, 2020). A project that degrades a relationship has consumed an asset, even if no accounting entry records the consumption. Materiality runs in two directions. Following the #double_materiality principle, an effect belongs in the appraisal if it is likely to affect enterprise value, or if the project materially affects people or the environment, or both (Baumuller and Sopp, 2022). This commitment is what prevents the framework from collapsing back into a risk register. Causal claims require a counterfactual. Following the impact measurement tradition, no effect may be claimed as a project benefit unless the appraisal states what would have happened without the project and why the project changes it (Corvo, Pastore, Manti, and Iannaci, 2021; Busch and colleagues, 2021). A #theory_of_change is the minimum acceptable form of this statement. Projects sit inside systems, and systems push back. Following the systems perspective advanced by Grewatsch, Kennedy, and Bansal, complex social and environmental problems cannot be decomposed into independent variables to be optimised one at a time, and interventions produce feedback, delay, and displacement (Grewatsch, Kennedy, and Bansal, 2023). #systems_thinking implies that the appraisal should look for rebound effects and for shifted burdens rather than only for direct effects. Two further theories explain organisational behaviour rather than justify the framework. #legitimacy_theory suggests that firms adopt visible practices to maintain their social licence, which predicts that impact appraisal will be adopted symbolically unless it is tied to consequences. #institutional_theory predicts convergence on whatever form regulators and peers reward, which predicts that impact appraisal will drift towards whatever is easiest to disclose. Both predictions are uncomfortable, and both should be treated as design constraints rather than as pessimism. A framework that cannot survive symbolic adoption and institutional drift is not a serious framework. From these commitments follows the central construct of the article, #integrated_value. Integrated value is the total change in wellbeing attributable to a project, distinguished by who receives it, expressed in the most credible unit available for each effect, and presented to decision makers with its uncertainty intact. Three features of this definition deserve emphasis. It is attributional, so it demands a counterfactual. It is distributional, so it refuses to net a gain to a wealthy group against a loss to a poor one without saying so. And it is plural in its units, so it does not require monetisation where monetisation would be dishonest. 5. Method This is a conceptual article built on an #integrative_review, a method appropriate where the aim is to synthesise evidence across disconnected fields and to build a framework rather than to test a hypothesis. Search and selection. Literature was gathered from the five streams identified above. Priority was given to work published from 2020 onwards, with older work admitted only where it is foundational and has not been superseded. Sources were drawn from peer reviewed journals in management, accounting, finance, and project studies, from academic books published by university and major commercial presses, and from a small number of authoritative public appraisal documents where they represent the state of practice rather than opinion. #inclusion_criteria. A source was included if it met at least one of four tests: it theorises the relationship between corporate activity and societal outcomes; it evaluates the reliability of an impact or sustainability metric; it examines how projects or investments are appraised in practice; or it provides evidence on the behavioural and organisational conditions under which appraisal practices change behaviour. Advocacy pieces without evidence, vendor material, and consultancy reports without disclosed method were excluded. #thematic_synthesis. Sources were coded against three questions: what does this tell us about what should be measured, what does it tell us about how it can be measured, and what does it tell us about why measurement does or does not change decisions. Codes were then clustered into the metric families, barriers, and enablers presented in Section 6. Framework construction. The framework in Section 7 follows a #design_science logic. It begins from a practical problem, derives design requirements from the evidence, proposes an artefact in the form of a process, and states the conditions under which the artefact would be judged successful. It has not yet been tested in the field. This is stated plainly rather than hidden, and Section 11 sets out how it should be tested. Illustrations. Section 8 presents three stylised project types. These are composites constructed to demonstrate how the framework operates. They are not case studies of real organisations, no real firm is described, and no empirical claim is made about any specific project. They should be read as worked examples, in the way that a textbook works an example, and not as evidence. 6. Analysis: Metrics, Barriers, and Enablers 6.1 Four families of societal impact metrics The synthesis identifies four #metric_families available to a project team. They differ in what they measure, how much work they require, and how far they can be trusted. Family one: activity and output indicators. These count what the project does. Hours of training delivered, tonnes of waste diverted, number of suppliers audited, number of local firms contracted. They are cheap, fast, and verifiable, which explains their popularity. Their weakness is decisive. They say nothing about whether anything changed for anybody. A firm can deliver ten thousand training hours to people who forget everything within a month. Output indicators belong in a business case as evidence of delivery, never as evidence of impact. Family two: ratings and scores. These aggregate many indicators into a composite judgement, usually by a third party. Their attraction is comparability across firms and time. The evidence reviewed above shows that the attraction is largely false, because raters disagree substantially and disagreement grows rather than shrinks with disclosure (Berg, Kolbel, and Rigobon, 2022; Christensen, Serafeim, and Sikochi, 2022). Ratings are also firm level and backward looking, so they cannot answer a project level and forward looking question. Their legitimate use in appraisal is narrow: as a screening signal, as a check on external perception, and as a proxy for how a decision may be received. They should never carry weight in the decision itself. Family three: #outcome_metrics with a counterfactual. These measure change in the state of the world attributable to the project. Household income of participating farmers relative to a comparison group. Rate of hospital readmission among users of a device relative to standard care. Groundwater level in the catchment relative to the modelled level without abstraction. Building these metrics requires a theory of change, a baseline, a comparison of some kind, and time. They are the only family that measures what the appraisal claims to be about. Their cost is real, and proportionality matters: a full comparison group design is justified for a flagship programme and absurd for a small procurement change. Family four: monetised impact. These convert outcomes into currency using shadow prices, damage costs, or valuation coefficients, allowing the result to be added to or subtracted from financial figures. The impact weighted accounts programme demonstrates that this can be done at scale and that it changes conclusions (Serafeim and Trinh, 2020; Freiberg, Park, Serafeim, and Zochowski, 2021). The Green Book tradition demonstrates that it can be done consistently if a public infrastructure of valuation coefficients exists (HM Treasury, 2022). #monetisation makes trade offs explicit, which is its great virtue, since a decision maker forced to see that a project earns twenty million and destroys thirty million of health is confronted rather than reassured. The dangers of #shadow_pricing are equally clear. Coefficients embed contested ethical choices, particularly around the value of a life year, the treatment of future generations through the discount rate, and the weight given to harm falling on the poor. A monetised total also invites netting, so that a large gain to many can silently cancel a severe harm to few. The framework therefore adopts a rule of restraint. Monetise where credible published coefficients exist and where the outcome is well established. Otherwise, report the outcome in its own units and let the decision maker weigh it. Never present a single net figure without the components that produced it. The practical conclusion is that no family is sufficient alone. A serious appraisal uses outputs to check delivery, outcomes to establish impact, monetisation where it is honest, and ratings only as context. 6.2 Barriers to integration The review surfaces six barriers that recur across literatures. #data_quality. Impact data is often self reported, collected inconsistently, and unaudited. Supply chain data is the worst case, since the effects a firm most needs to see are furthest from the effects it directly controls. Kaplan and Ramanna's proposal for emissions accounting is precisely an attempt to solve this by replacing estimates with audited transfers between firms (Kaplan and Ramanna, 2021). #comparability. Without common definitions, two projects cannot be compared, and capital cannot be allocated between them. The divergence findings show that the market has not solved this problem and may not solve it soon (Berg, Kolbel, and Rigobon, 2022). Time horizon and discounting. Many societal effects arrive slowly. A commercial discount rate applied over thirty years reduces them to rounding errors. This is a technical choice with a moral consequence, and it should be surfaced in the appraisal rather than buried in a spreadsheet cell. #incentives. Project managers are promoted for delivering on time and on budget. Sponsors are rewarded for getting approval. Neither is rewarded for an honest impact estimate that weakens their case. Silvius and Schipper's finding that personal responsibility and organisational support drive sustainability behaviour, more than formal procedures, points to where the leverage lies (Silvius and Schipper, 2020). Flyvbjerg's work on strategic misrepresentation explains what happens when the leverage is absent (Flyvbjerg, 2021). Capability. Most project teams contain nobody trained in impact evaluation. Asking a finance analyst to build a theory of change without support produces a decorated cash flow model. #impact_washing. The reputational reward for claiming impact currently exceeds the cost of failing to deliver it. Until claims are assured, this asymmetry will corrupt measurement, and the greenwashing literature suggests detection is genuinely hard (de Freitas Netto and colleagues, 2020). 6.3 Enablers Against these barriers, the literature points to conditions that make integration more likely to hold. Governance with teeth. Impact evidence must be required at the same gate as financial evidence, reviewed by the same committee, and capable of stopping a project. If impact appears in an appendix, it will be ignored. #assurance. Impact claims that pass into decisions should be subject to internal audit and, for material claims, external assurance, on the same principle that financial figures are audited. #stakeholder_engagement as method, not public relations. Affected people are the cheapest and most accurate source of information about what a project will do to them. Engagement that happens after design is complete is a communication exercise. Engagement that shapes the theory of change is an evidence gathering exercise. Proportionality. A rule that every project must build a comparison group design will be ignored within a year. A tiered rule, in which depth of analysis scales with size and severity of potential harm, can survive. #capacity_building. Training project professionals in the basics of theory of change, counterfactual reasoning, and valuation is a small investment with large returns, and it is the practical answer to the paradox that Sabini and Alderman describe (Sabini and Alderman, 2021). Reference class discipline. Impact forecasts, like cost forecasts, should be checked against what similar projects actually achieved, and organisations should retain the data that makes such checks possible (Flyvbjerg and Gardner, 2023). 7. The Integrated Project Value Framework The framework has five stages and one loop. It is designed to sit inside a conventional stage gate process rather than to replace it, because a framework that requires organisational redesign before it can be used will never be used. Stage 1: Scoping and materiality The first stage asks a single question. Who is affected by this project, and how? The team constructs a map of affected groups, including those with no commercial relationship to the firm, such as neighbouring communities, downstream water users, future users of a shared resource, and workers deep in the supply chain. For each group, the team lists plausible effects, positive and negative, direct and indirect. Each candidate effect is then assessed against two axes, following the double materiality logic. The first axis asks how much the effect could matter to the firm, in cash, risk, or licence to operate. The second asks how much the project could matter to the affected group, in severity, scale, and reversibility. The output is a #materiality_matrix. Effects that score high on either axis enter the appraisal. Effects that score low on both are recorded and set aside, with the reasoning documented so that a reviewer can challenge it. Two rules protect the integrity of this stage. Severity trumps probability where harm is irreversible, so a small chance of destroying an aquifer is material even if the expected value looks modest. And the map must be tested with at least one voice from outside the project team, because a team that maps its own stakeholders will map the ones it already knows. Stage 2: Theory of change and baseline For every material effect, the team writes a short causal chain: the project does X, which leads to Y, which changes Z for group G, within timeframe T, under assumptions A. The chain must include the assumptions, because the assumptions are where the argument is weakest and therefore where a reviewer should look first. The team then establishes a #baseline, which is the state of the world today, and a counterfactual, which is the expected state of the world without the project. These are different things and confusing them is the commonest error in impact claims. A firm that builds a solar plant in a grid that would otherwise have added solar anyway has a baseline of high emissions and a counterfactual of low ones, and its true impact is close to zero even though its output metric looks excellent. At this stage the team also identifies possible rebound and displacement effects. Efficiency gains that reduce cost can increase consumption. Local hiring can displace other local employers. A systems view expects these and looks for them. Stage 3: Measurement and valuation Each material effect is assigned to a metric family using a simple decision rule. If the outcome is well established and a credible published valuation coefficient exists, monetise it and carry it into the financial model as an explicit line, never blended into revenue or cost. Greenhouse gas emissions, air quality effects, and certain health outcomes usually qualify. If the outcome can be measured but not credibly valued, quantify it in its natural units and carry it into the decision as a named quantity. Litres of water abstracted, jobs sustained beyond two years, injury rate among contract workers. If the outcome can be neither measured nor valued with any confidence, describe it, state why it cannot be quantified, and carry it forward as a qualitative flag. Do not silently drop it. The Green Book tradition insists on exactly this discipline, and it is the single most important protection against the tyranny of what is easy to count (HM Treasury, 2022). Uncertainty is expressed as ranges rather than point estimates, and the appraisal states which assumption the conclusion is most sensitive to. Where the project relies on an optimistic impact assumption, the team should ask what evidence would falsify it and how much it would cost to obtain that evidence before committing capital. Stage 4: Decision rules This is the stage most frameworks avoid, and avoiding it is what turns impact analysis into decoration. The framework proposes a three test structure. Test one, the harm floor. No project proceeds if it causes severe and irreversible harm to a vulnerable group, regardless of its financial return. This is a constraint, not a trade off, and it exists because some trade offs should not be offered. Test two, the integrated value test. The project is assessed on #impact_adjusted_net_present_value, defined as conventional net present value plus monetised benefits to society minus monetised costs to society, with each component shown separately and never netted into a single headline without its parts. A project whose commercial NPV is positive only because it exports costs to others fails this test even though it passes the financial one. Test three, the qualitative panel. Effects that could not be monetised are presented as a structured set of named quantities and flags, and the approving body must record its judgement on each. This is deliberately not automated. The purpose of the framework is to force judgement into the open, not to remove it. Requiring a written rationale creates the audit trail that Bebchuk and Tallarita rightly demand of any stakeholder claim (Bebchuk and Tallarita, 2020). The #decision_rules also require that where impact evidence changes the decision, the change is recorded. An organisation that has never rejected or redesigned a project on impact grounds has not implemented the framework, whatever its documents say. Stage 5: Monitoring, verification, and benefits realisation Approval is not the end of appraisal. The impacts claimed at the gate become commitments to be tracked, with named owners, defined measures, and reporting dates, in exactly the way that financial benefits are tracked under a #benefits_realisation regime. Material claims are subject to assurance. Where the claimed impact does not materialise, the organisation must say so. The loop: learning and reference classes The final element is a #learning_loop. Realised impacts are compared with forecast impacts, and the gaps are stored in a database organised by project type. Over time this becomes the organisation's reference class, and future forecasts are adjusted against it. This is the impact equivalent of reference class forecasting for cost, and it is the only reliable defence against optimism bias in impact claims (Flyvbjerg, 2021; Flyvbjerg and Gardner, 2023). Without the loop, the framework is a better looking business case. With the loop, it is a system that gets less wrong over time. Proportionality tiers To keep the framework usable, three tiers are proposed. Tier one, light touch. Small projects with low potential harm. Stages one and four only, in a single page. Perhaps an hour of work. Tier two, standard. Most capital projects. All five stages, with quantified outcomes, monetisation where coefficients exist, and internal review. Tier three, deep. Large projects, projects affecting vulnerable populations, or projects with irreversible environmental consequences. All five stages, external stakeholder input, external assurance of material claims, and a formal counterfactual design. The tier is set by a short screening test at project initiation, and the assignment itself is reviewable, because the incentive to under classify will otherwise be irresistible. 8. Illustrative Applications The following are stylised examples, constructed to show the mechanics of the framework. They describe no real organisation and make no empirical claim. 8.1 A grid scale renewable energy project A utility evaluates a #renewable_energy plant. The conventional case is straightforward: capital cost, expected generation, tariff, operating cost, positive NPV. Under the framework, Stage 1 identifies affected groups beyond customers and shareholders: landholders, a nearby community reliant on grazing, workers in the fossil plant the project may displace, and the wider population affected by emissions. Stage 2 forces the crucial question. What would the grid have done without this plant? If the counterfactual is a gas plant, the emissions benefit is large. If the counterfactual is a different renewable project by a competitor, the benefit is small, and the true contribution is speed rather than substitution. This distinction is invisible in a conventional case and decisive in an honest one. Stage 3 monetises avoided emissions using a published damage cost, and quantifies land use change in hectares and community grazing loss in households affected, both of which resist honest valuation. It also flags the transition cost to fossil plant workers as an outcome the firm can influence. Stage 4 shows a project that passes the harm floor, has a strongly positive impact adjusted NPV driven by avoided emissions, and carries two qualitative burdens. The approving committee is required to record how it will address the grazing loss and the worker transition, which turns a public relations problem into a design input. In several plausible variants of this example, the project is approved with a different land footprint than originally proposed, which is what integration looks like when it is working. 8.2 A #digital_health product A firm evaluates a monitoring product sold to hospitals. The commercial case rests on subscription revenue. Stage 1 identifies patients, clinicians, and health systems as affected groups, none of whom appear in the revenue line. Stage 2 produces a theory of change: the device detects deterioration earlier, which prompts earlier intervention, which reduces readmission, which improves patient outcomes and reduces system cost. Every link is an assumption and each can fail. The device may generate alerts that clinicians ignore. Earlier detection may not change treatment. The theory of change makes these failure points visible before launch rather than after. Stage 3 measures readmission against a comparison group, which is feasible in a health setting and is the difference between a claim and a fact. Where evidence is thin, the framework requires the firm to say so and to fund a study rather than to assert an outcome in a marketing deck. It also asks about harms that a commercial case would never raise, such as alert fatigue among nurses and the risk that the product performs worse for patient groups underrepresented in its training data. Stage 4 may show a project with modest commercial NPV and large societal value, which is an argument for pricing that shares the value rather than for abandoning the project. It may equally show large commercial NPV built on an unproven clinical claim, which is a reputational and regulatory liability disguised as a growth opportunity. 8.3 A #supply_chain cost reduction programme A manufacturer evaluates a programme to move sourcing to lower cost suppliers. The conventional case is a clean saving with a fast payback. Stage 1 identifies workers at incumbent and new suppliers, communities dependent on the incumbent, and the firm's own customers. Stage 2 asks what happens to wages, hours, and safety at the new suppliers, and whether the saving is a genuine efficiency or a transfer from workers to the firm. This distinction is the heart of the matter and is entirely absent from a conventional appraisal, which sees only a lower unit price. Stage 3 quantifies #local_employment lost in the incumbent region, monetises nothing that cannot be honestly monetised, and flags labour standards risk with an explicit likelihood and severity. Stage 4 applies the harm floor. If the saving depends on conditions the firm would not accept in its own operations, the project is redesigned or refused, not discounted. The lesson of this example is uncomfortable and therefore worth stating. Integration will sometimes reduce reported financial returns, because some reported returns were never returns at all. They were transfers that the accounting system was not built to see. 9. Discussion 9.1 Theoretical implications The framework contributes to three conversations. To the purpose literature it offers mechanics, which is what George and colleagues identify as missing when they observe that purpose research is rich in claims and poor in evidence (George and colleagues, 2023). Purpose that cannot change a capital allocation decision is not purpose in any operational sense. To the ESG literature it offers a way out of the measurement impasse. The divergence findings are often read as an argument for abandoning measurement. They are better read as an argument for abandoning outsourced measurement (Berg, Kolbel, and Rigobon, 2022; Edmans, 2023). A firm cannot buy its way to knowing whether its projects help or harm. It has to do the work at the level where the effects are produced, which is the project. To project studies it offers an answer to the question of what success means, which Ika and Pinto raise and which cannot be answered without an appraisal method (Ika and Pinto, 2022). If success includes societal value, then appraisal must estimate societal value, and if appraisal does not, the redefinition of success remains a slogan. 9.2 Practical implications For sponsors, the framework changes what a good business case looks like. It is no longer a document that proves a project should proceed. It is a document that states honestly what the project will do to whom, with its uncertainty visible. The confident single number is a warning sign, not a strength. For project professionals, the framework requires new skills, chiefly counterfactual reasoning and the ability to write a defensible theory of change. This is a training problem before it is a systems problem, and it is solvable. #training in these methods is far cheaper than the cost of one project stopped by a community that the appraisal never mapped. For boards and investment committees, the framework demands something harder than approval or rejection. It demands recorded judgement on effects that resist quantification, and it removes the possibility of hiding behind a number. #decision_usefulness improves not because the framework produces certainty but because it produces clarity about what is being traded away. For governance functions, the framework has implications for #project_governance and #capital_allocation. Gate criteria must be rewritten, impact owners must be named, and assurance must extend to material impact claims. Where impact claims influence executive remuneration, they must be assured, because an unassured metric attached to a bonus will be gamed. 9.3 The trade off problem The framework does not dissolve #trade_offs and does not pretend to. A project can raise living standards and damage a river. The framework's contribution is to make the trade explicit, to name who gains and who loses, and to require that the decision maker own the choice. This is less satisfying than a formula and more honest than one. Freeman and colleagues make the point that the presence of tension is not a defect in stakeholder thinking but its central subject matter (Freeman, Phillips, and Sisodia, 2020). There is also a distributional question that no aggregate can answer. A monetised total treats a dollar of benefit to a wealthy shareholder as equal to a dollar of cost to a poor household. Public appraisal traditions handle this with distributional weights, and commercial appraisal will eventually have to confront it. At minimum, the framework requires that gains and losses be reported by group, so that a decision maker who chooses to net them does so knowingly. 9.4 Regulation and the limits of voluntarism Firms that integrate impact honestly may find themselves at a short term disadvantage against firms that do not, because honesty costs money and imposes constraints. This is a collective action problem, and collective action problems are usually solved by rules rather than by virtue. #regulation that standardises definitions, mandates disclosure of material impacts, and, in time, prices the largest externalities is therefore not the enemy of integrated appraisal but its precondition. Firms that build the capability early are better placed when the rules arrive, which is a defensible commercial argument for doing the work before it is required. The convergence of sustainability reporting standards has already begun to create the shared vocabulary that appraisal needs, and appraisal practice should be read as the internal counterpart of external reporting rather than as a separate initiative (Adams and Abhayawansa, 2022; Baumuller and Sopp, 2022). 10. Limitations Several #limitations should be stated plainly. The framework is conceptual and has not been tested empirically. It is derived from evidence but is not itself evidence. The literature reviewed is dominated by work from Europe and North America, and the valuation coefficients on which monetisation depends are largely built on high income country data. Applying them elsewhere without adjustment would be a serious error, and the article does not solve this problem. The framework asks organisations to do more work at appraisal, and the assumption that better decisions justify the cost is plausible but unproven. #generalisability across sectors is also untested. Appraisal in pharmaceuticals, mining, and consumer software involves very different effect types and evidence standards. Finally, the framework can be captured. An organisation determined to approve a project can produce a theory of change that reaches the desired conclusion. The defences proposed here, assurance, external voices, recorded judgement, and a learning loop that compares forecasts with outcomes, raise the cost of capture but do not eliminate it. No method can protect a decision from a decision maker who has already decided. 11. Future Research Five directions follow directly. Field testing. The framework should be applied in real stage gate processes and studied through action research, with attention to whether it changes decisions rather than only documents. #longitudinal_studies. Forecast impacts should be compared with realised impacts across a portfolio, over years, to build the reference classes the framework depends on and to measure optimism bias in impact claims as Flyvbjerg has measured it in cost claims. Valuation infrastructure. Research is needed on credible, context adjusted coefficients for the effects that commercial projects most often produce, and on the sensitivity of conclusions to those coefficients. Behaviour and incentives. Experimental work should test how decision makers actually use plural information, and whether the presence of unquantified flags changes choices or is quietly ignored in favour of the number. #open_data and comparability. Work is needed on shared definitions at the project level, and on whether #open_data on realised project impacts could function as the reference class infrastructure that no single firm can build alone. Underlying all five is a single empirical question that the field has not answered: does integrating societal impact into appraisal produce better outcomes, for society and for firms, than the alternatives? #future_research should treat that question as open, because assuming the answer is exactly the intellectual habit that produced the ESG disappointments this article has described. 12. Conclusion The business case is not a neutral technology. It is a statement about what counts, and for most of a century it has counted a narrow slice of what a project actually does. That narrowness was tolerable when the effects it excluded fell on people with no power to respond. It is not tolerable now, and the evidence suggests it is not even profitable, since the excluded effects return as delays, litigation, regulation, lost talent, and lost trust. This article has argued that the answer is neither to abandon financial discipline nor to bolt a sustainability appendix onto an unchanged model. It is to widen the appraisal itself, holding impact claims to the same evidential standard as revenue claims, and refusing the two easy escapes: the claim that impact cannot be measured, and the claim that it can be reduced to a score. The Integrated Project Value framework offers a way to do this. Map who is affected. State how the project changes their lives and what would have happened otherwise. Measure what can be measured, value what can honestly be valued, and describe the rest rather than deleting it. Decide with the harm floor intact and the trade offs visible. Then track what actually happened and let it correct the next forecast. None of this is exotic. Public appraisal has done versions of it for decades, and impact evaluation has refined the methods for longer. What is new is the proposal that a commercial firm should do it as a matter of routine, inside the process that allocates its capital, and should accept the consequences when the analysis says no. #redefining_the_business_case is finally a question of #value_creation rather than of reporting. A firm that knows what its projects do to the world is not merely a more #responsible_management proposition. It is a better informed one, and in a world where the excluded costs are coming back, better informed is the same thing as better run. Hashtags #Societal_Impact_Metrics #Business_Case_Redefined #Project_Appraisal #Impact_Measurement #Sustainable_Project_Management #Double_Materiality #Stakeholder_Value #ESG_Critique #Impact_Weighted_Accounting #Corporate_Purpose #Responsible_Investment #Social_Value #Integrated_Reporting #Capital_Allocation #Management_Research References Adams, C. A., and Abhayawansa, S. (2022). Connecting the COVID-19 pandemic, environmental, social and governance investing and calls for harmonisation of sustainability reporting. Critical Perspectives on Accounting, 82, 102309. Barnett, M. L., Henriques, I., and Husted, B. W. (2020). Beyond good intentions: Designing CSR initiatives for greater social impact. Journal of Management, 46(6), 937 to 964. Baumuller, J., and Sopp, K. (2022). Double materiality and the shift from non financial to European sustainability reporting: Review, outlook and implications. Journal of Applied Accounting Research, 23(1), 8 to 28. Bebchuk, L. A., and Tallarita, R. (2020). The illusory promise of stakeholder governance. Cornell Law Review, 106(1), 91 to 178. Berg, F., Kolbel, J. F., and Rigobon, R. (2022). Aggregate confusion: The divergence of ESG ratings. Review of Finance, 26(6), 1315 to 1344. Busch, T., Bruce-Clark, P., Derwall, J., Eccles, R., Hebb, T., Hoepner, A., Klein, C., Krueger, P., Paetzold, F., Scholtens, B., and Weber, O. (2021). Impact investments: A call for reorientation. SN Business and Economics, 1(2), 33. Christensen, D. M., Serafeim, G., and Sikochi, A. (2022). Why is corporate virtue in the eye of the beholder? The case of ESG ratings. The Accounting Review, 97(1), 147 to 175. Corvo, L., Pastore, L., Manti, A., and Iannaci, D. (2021). Mapping social impact assessment models: A literature overview for a future research agenda. Sustainability, 13(9), 4750. Dasgupta, P. (2021). The Economics of Biodiversity: The Dasgupta Review. London: HM Treasury. de Freitas Netto, S. V., Sobral, M. F. F., Ribeiro, A. R. B., and Luz Soares, G. R. (2020). Concepts and forms of greenwashing: A systematic review. Environmental Sciences Europe, 32(1), 19. Edmans, A. (2020). Grow the Pie: How Great Companies Deliver Both Purpose and Profit. Cambridge: Cambridge University Press. Edmans, A. (2023). The end of ESG. Financial Management, 52(1), 3 to 17. Flyvbjerg, B. (2021). Top ten behavioral biases in project management: An overview. Project Management Journal, 52(6), 531 to 546. Flyvbjerg, B., and Gardner, D. (2023). How Big Things Get Done. New York: Currency. Freeman, R. E., Phillips, R., and Sisodia, R. (2020). Tensions in stakeholder theory. Business and Society, 59(2), 213 to 231. Freiberg, D., Park, D. G., Serafeim, G., and Zochowski, T. R. (2021). Corporate environmental impact: Measurement, data and information. Harvard Business School Working Paper 20-098. George, G., Haas, M. R., McGahan, A. M., Schillebeeckx, S. J. D., and Tracey, P. (2023). Purpose in the for profit firm: A review and framework for management research. Journal of Management, 49(6), 1841 to 1869. Gillan, S. L., Koch, A., and Starks, L. T. (2021). Firms and social responsibility: A review of ESG and CSR research in corporate finance. Journal of Corporate Finance, 66, 101889. Grewatsch, S., Kennedy, S., and Bansal, P. (2023). Tackling wicked problems in strategic management with systems thinking. Strategic Organization, 21(3), 721 to 732. Hart, O., and Zingales, L. (2022). The new corporate governance. University of Chicago Business Law Review, 1(1), 195 to 216. Heeb, F., Kolbel, J. F., Paetzold, F., and Zeisberger, S. (2023). Do investors care about impact? The Review of Financial Studies, 36(5), 1737 to 1787. HM Treasury (2022). The Green Book: Central Government Guidance on Appraisal and Evaluation. London: HM Treasury. Ika, L. A., and Pinto, J. K. (2022). The re meaning of project success: Updating and recalibrating for a modern project management. International Journal of Project Management, 40(7), 835 to 848. Ika, L. A., Soderlund, J., Munro, L. T., and Landoni, P. (2020). Cross learning between project management and international development: Analysis and research agenda. International Journal of Project Management, 38(8), 548 to 558. Kaplan, R. S., and Ramanna, K. (2021). Accounting for climate change. Harvard Business Review, 99(6), 120 to 131. Kolbel, J. F., Heeb, F., Paetzold, F., and Busch, T. (2020). Can sustainable investing save the world? Reviewing the mechanisms of investor impact. Organization and Environment, 33(4), 554 to 574. Locatelli, G., Ika, L., Drouin, N., Muller, R., Huemann, M., Soderlund, J., Geraldi, J., and Clegg, S. (2023). A manifesto for project management research. European Management Review, 20(1), 3 to 17. Mayer, C. (2021). The future of the corporation and the economics of purpose. Journal of Management Studies, 58(3), 887 to 901. Mazzucato, M. (2021). Mission Economy: A Moonshot Guide to Changing Capitalism. London: Allen Lane. Pastor, L., Stambaugh, R. F., and Taylor, L. A. (2021). Sustainable investing in equilibrium. Journal of Financial Economics, 142(2), 550 to 571. Pedersen, L. H., Fitzgibbons, S., and Pomorski, L. (2021). Responsible investing: The ESG efficient frontier. Journal of Financial Economics, 142(2), 572 to 597. Pucker, K. P. (2021). Overselling sustainability reporting. Harvard Business Review, 99(3), 134 to 143. Sabini, L., and Alderman, N. (2021). The paradoxical profession: Project management and the sustainability of built environments. Project Management Journal, 52(4), 379 to 393. Serafeim, G. (2022). Purpose and Profit: How Business Can Lift Up the World. New York: HarperCollins Leadership. Serafeim, G., and Trinh, K. (2020). A framework for product impact weighted accounts. Harvard Business School Working Paper 20-076. Silvius, A. J. G., and Schipper, R. (2020). Exploring variety in factors that stimulate project managers to address sustainability issues. International Journal of Project Management, 38(6), 353 to 367. van Zanten, J. A., and van Tulder, R. (2021). Improving companies impacts on sustainable development: A nexus approach to the Sustainable Development Goals. Business Strategy and the Environment, 30(8), 3703 to 3720. Zwikael, O., and Meredith, J. R. (2021). Evaluating the success of a project and the performance of its leaders. IEEE Transactions on Engineering Management, 68(6), 1745 to 1757.

The COVID-19 pandemic forced a sudden and very large experiment in how projects are planned, staffed, monitored and delivered. Between 2020 and 2022, millions of knowledge workers moved from shared offices to home offices almost overnight. What began as an emergency response has now settled into a durable working arrangement, and the question facing scholars and practitioners is no longer whether remote and hybrid project work is possible, but how the behaviour of project teams and project managers has changed because of it. This article presents an integrative review of peer reviewed research published between 2021 and 2025 on #post_pandemic_project_management, with a specific focus on #remote_team_administration and #digital_collaboration. Using a structured search and thematic synthesis of empirical and conceptual studies drawn from management, information systems, organisational psychology and software engineering, the review identifies six behavioural shifts that consistently appear across the evidence base. These are: the movement of coordination from spontaneous conversation to written and #asynchronous_communication; the narrowing and hardening of #collaboration_networks; the reconstruction of trust from relationship based trust to task based and evidence based trust; the redistribution of control from presence based supervision to #outcome_based_evaluation; the migration of project knowledge from human memory to shared digital repositories; and the emergence of new wellbeing risks connected to boundary loss and meeting overload. The review then proposes a Behavioural Shift Framework that links these six changes to established theory in #work_design, #media_richness and coordination research, and offers an agenda for practice, for teaching and for future study. The central argument is that remote and hybrid project management is not simply office project management delivered through a screen. It is a different behavioural system, with different default behaviours, different failure modes and different competency requirements. Project organisations that treat it as a technology problem tend to underperform, while those that treat it as a behaviour and design problem tend to do better. Keywords: post-pandemic project management; remote teams; hybrid work; digital collaboration; behavioural change; virtual team trust; asynchronous coordination; project governance; agile practice; future of work 1. Introduction 1.1 Background to the problem Project management has always been a social activity dressed up in technical clothing. Schedules, budgets, risk registers and work breakdown structures are the visible surface of the discipline. Underneath that surface sits a set of human behaviours: people asking each other questions, people noticing that something is wrong, people deciding whether to raise a concern, people trusting a colleague enough to hand over a task. For most of the history of the discipline, those behaviours were supported by a physical office. A project manager could walk past a desk and see that a team member looked stressed. A tester could turn around and ask a developer a quick question. A sponsor could be caught in a corridor for two minutes and give an informal approval. The pandemic removed that physical support structure with almost no notice. Bick, Blandin and Mertens documented how quickly the share of workers operating from home rose in the United States after the outbreak, and how a large part of that increase persisted afterwards rather than snapping back to the previous baseline (Bick et al., 2023). Barrero, Bloom and Davis argued early on that the shift would prove sticky, because both employees and employers had discovered value in it that they had not previously priced in (Barrero et al., 2021). Several years later, that prediction looks broadly correct. #hybrid_work has become the normal arrangement for a large share of graduate level knowledge work in Europe and North America, and fully distributed teams are now common in software, consulting, research and design. For the project management community, this raises a specific and practical question. If the behaviours that made projects work were partly produced by shared physical space, and that shared space has been reduced or removed, then what has happened to those behaviours? Have they been replaced by equally effective digital substitutes, have they degraded, or have they been transformed into something new that we do not yet measure properly? 1.2 Why the behavioural angle matters Much of the early literature on pandemic era work focused on productivity. This was understandable, because productivity is measurable and because senior managers were anxious about it. But productivity is an outcome, not a mechanism. It tells us what happened without telling us why. A team can appear productive on a dashboard while its informal knowledge sharing quietly collapses, and the cost of that collapse may only show up two years later in poor onboarding, weak innovation and higher turnover. The behavioural angle is therefore not a softer alternative to the productivity question. It is the layer underneath it. Yang and colleagues showed, using a very large dataset of internal communication and calendar records, that firm wide remote work made collaboration networks more static and more siloed, and pushed communication towards asynchronous channels and away from richer synchronous ones (Yang et al., 2022). That is a behavioural finding with structural consequences. It suggests that the pandemic did not simply move existing conversations online. It changed which conversations happened at all, and between whom. Similarly, the randomised controlled trial reported by Bloom, Han and Liang found that a hybrid schedule of two days at home per week improved job satisfaction and cut quit rates by roughly a third, without damaging performance grades over the following two years (Bloom et al., 2024). Again, the interesting part is behavioural. It is not simply that hybrid work "works". It is that the behaviour of staying with an employer, and the behaviour of managers when they rate performance, both changed in measurable ways. 1.3 The gap in the literature Three gaps motivate this review. First, the evidence base is scattered across disciplines that rarely read each other. Software engineering researchers have produced detailed studies of distributed #agile_practices. Organisational psychologists have produced strong work on wellbeing and #work_design. Economists have produced careful causal studies of productivity and retention. Information systems researchers have studied technology adoption and affordances. Project management as a field sits at the crossroads of all of these, but there are few reviews that pull the threads together with the project as the unit of analysis. Second, much of the early work was written in emergency conditions, during 2020 and 2021, when remote work was involuntary, when schools were closed and when stress levels were unusually high. Findings from that period cannot be transferred directly to the present situation, in which remote and hybrid work is largely voluntary and organisationally planned. Smite and colleagues made this distinction explicit, describing the move from forced working from home to voluntary working from anywhere as two separate revolutions with different psychological character (Smite, Moe, Klotins and Gonzalez-Huerta, 2023). Third, the practitioner literature has raced ahead of the research literature. Guidance about hybrid meetings, asynchronous documentation and digital tooling is abundant, but a good deal of it is based on the experience of a single company or on consultancy opinion rather than on systematic evidence. Students in particular need a synthesis that separates what is reasonably well established from what is merely popular. 1.4 Research questions This article addresses four questions. What behavioural changes in project teams and project managers are reported consistently in peer reviewed research published between 2021 and 2025? Through which mechanisms do these changes affect coordination, trust, knowledge flow and control in projects? What existing theory best explains these mechanisms, and where does existing theory fall short? What do the findings imply for project practice, for project management education and for future research? 1.5 Contribution and structure The contribution of this article is threefold. It provides a cross disciplinary synthesis of the post-pandemic evidence with the project as the focal unit. It proposes a Behavioural Shift Framework that organises the findings into six interlocking changes rather than a long unstructured list of challenges. And it translates the synthesis into concrete implications for project governance, for competency development and for research design. The remainder of the article is organised as follows. Section 2 reviews the literature across six thematic streams. Section 3 sets out the theoretical foundations and presents the framework. Section 4 explains the review method. Section 5 reports the findings. Section 6 discusses them. Section 7 draws out implications. Section 8 states limitations, Section 9 proposes a research agenda, and Section 10 concludes. 2. Literature Review 2.1 The pre-pandemic starting point Distributed project work is not new. Global software development, offshore engineering and multinational construction programmes have used dispersed teams for decades, and a substantial body of research on virtual teams already existed before 2020. What was new after 2020 was scale, speed and involuntariness. Vartiainen, reviewing the telework and remote work literature, notes that the field had long documented both the flexibility benefits and the coordination costs of distance, but that these were usually studied in settings where distributed work was a deliberate strategic choice made by a minority of firms (Vartiainen, 2021). The pandemic broke that assumption. Distance stopped being a strategic variable and became an environmental constraint. This matters because a great deal of pre-2020 advice assumed that organisations could select for remote readiness, could invest gradually and could keep a physical hub as an anchor. None of those assumptions held during the emergency period, and only some of them have returned. The project management profession was itself already in motion. The seventh edition of the widely used body of knowledge published by the Project Management Institute shifted from a process based structure to a principles and value delivery structure, giving more explicit attention to team behaviour, stewardship, tailoring and systems thinking (Project Management Institute, 2021). That change was not caused by the pandemic, but it made the profession's formal guidance more compatible with the behavioural questions the pandemic then forced onto the agenda. 2.2 Productivity, retention and the contested evidence The most publicly visible strand of the literature concerns whether remote work helps or harms output. The evidence is genuinely mixed, and students should be careful not to accept a single headline. On the positive side, Choudhury, Foroughi and Larson studied a work-from-anywhere transition at a large public organisation and found a measurable increase in individual output after workers gained geographic flexibility, alongside benefits related to reduced commuting and greater location choice (Choudhury et al., 2021). Bloom and colleagues, in a randomised trial that is methodologically among the strongest studies in this space, found that hybrid working improved satisfaction and #retention substantially while leaving performance grades unchanged (Bloom et al., 2024). Retention effects were strongest for non managers, for female employees and for those with long commutes. On the more cautious side, Gibbs, Mengel and Siemroth analysed detailed personnel and analytics data for information technology professionals and reported that total hours worked rose during the shift to home working while output did not rise in proportion, which implies a fall in productivity per hour (Gibbs et al., 2023). Yang and colleagues found that collaboration became more siloed, which is not directly a productivity measure but which points towards long term costs in #innovation and #knowledge_sharing (Yang et al., 2022). The apparent contradiction dissolves once we stop asking whether remote work is good or bad and start asking under what conditions and for which behaviours. Independent, well specified, individually owned tasks appear to survive distance well. Interdependent, ambiguous, negotiation heavy tasks appear to suffer. Since project work is by definition full of interdependence and ambiguity, this distinction is central to our topic. 2.3 Communication, coordination and collaboration networks The clearest and most replicated behavioural finding is a change in the shape of communication. Yang and colleagues documented that firm wide remote work made internal #collaboration_networks more static, meaning that people kept talking to the people they already knew and formed fewer new connections. It also increased the proportion of communication carried by asynchronous written channels rather than by richer synchronous channels such as scheduled or unscheduled calls (Yang et al., 2022). Both of these changes reduce the flow of information across group boundaries. De Souza Santos and Ralph, building a grounded theory of coordination in remote first and hybrid software teams, described how teams compensated by creating explicit coordination structures where implicit ones had previously been sufficient (de Souza Santos and Ralph, 2022). Where a co-located team could coordinate by overhearing, a distributed team must coordinate by writing. This raises #coordination_costs in the short term, but it also produces artefacts that can be reused, searched and audited. The loss of #weak_ties deserves particular emphasis. In a physical office, weak ties are cheap. You meet someone from another workstream at the coffee machine, and three weeks later you remember that they solved a similar problem. In a distributed environment, weak ties are expensive, because every interaction must be deliberately scheduled. Over time, teams drift towards #siloed_networks in which each cluster is well connected internally and poorly connected externally. For a project manager, siloed networks are dangerous, because cross workstream risks are precisely the risks that no single cluster can see. 2.4 Trust, psychological safety and team cohesion Trust in virtual settings has long been theorised as #swift_trust, in which team members extend provisional trust based on role, reputation and category membership rather than on personal history, and then adjust that trust based on early behaviour. The post-pandemic literature broadly supports this view but adds an important nuance. During 2020 and 2021, many teams were able to survive on trust that had been accumulated before the pandemic. People already knew their colleagues. The relational bank account was full, and it could be drawn down. The problems became visible later, when new members joined teams they had never met physically, and when the pre-existing stock of relational trust ran out. Agren, Knoph and Berntsson Svensson, studying agile software teams one year into the pandemic, reported exactly this pattern: technical practices adapted reasonably well, while social practices, informal learning and #team_cohesion degraded more slowly but more persistently (Agren et al., 2022). #psychological_safety is the second casualty worth noting. Speaking up is easier when a person can read the room. In a video call, particularly a large one with cameras off, the room cannot be read. Silence is ambiguous. A junior team member who is unsure whether a concern is worth raising receives no encouraging nod and no supportive glance, and so the concern is more likely to go unsaid. This is a behavioural mechanism with direct consequences for #risk_management, because most project risks are known to someone before they materialise. Wang and colleagues, analysing remote working through a work design lens, identified loneliness and ineffective communication as two of the four main challenges reported by remote employees, alongside work-home interference and procrastination, and identified social support, job autonomy, monitoring practices and workload as key moderating conditions (Wang et al., 2021). Their framework is useful because it treats these outcomes not as inevitable features of remote work but as results of how the work is designed. 2.5 Agile practices under distance Agile methods are unusually sensitive to distance, because they were explicitly designed around face to face interaction and co-located teams. The post-pandemic literature therefore provides a natural experiment in what happens when a method loses one of its founding assumptions. The findings are more encouraging than the theory would predict, but with clear caveats. Ceremonies transferred reasonably well. The #daily_stand_up survived in video form, planning sessions moved to shared digital boards, and retrospectives were often reported as being no worse and sometimes better, because quieter team members could contribute in writing. Ford and colleagues, studying software developers working from home, found a mixed picture in which many individual benefits coexisted with real difficulties in collaboration and in maintaining shared awareness (Ford et al., 2022). What transferred badly was the informal layer that agile assumes but does not specify: the ad hoc pairing, the whiteboard argument, the moment when someone notices confusion on a colleague's face and stops to explain. Conboy, Moe, Stray and Gundelsby argue that hybrid software development requires organisations to challenge assumptions that were previously invisible, including the assumption that co-location is the default and remote work the exception (Conboy et al., 2023). When the default flips, practices that were never written down have to be redesigned and made explicit. Smite and colleagues, examining the work policies that seventeen companies adopted after the emergency period, found wide variation and considerable tension between employee expectations and management expectations, and argued for flexibility rather than blanket mandates (Smite, Moe, Hildrum, Gonzalez-Huerta and Mendez, 2023). This tension between individual preference and organisational coordination need is one of the defining management problems of the current period. 2.6 Technology, monitoring and control The digital toolset expanded rapidly and then consolidated. Most project organisations now operate a broadly similar stack: a chat platform, a video platform, a work tracking system with #kanban_boards or backlogs, a shared document repository and some form of #project_dashboards. The interesting research question is not which tools are used, but what behaviour the tools produce. Two effects recur in the literature. The first is #digital_presenteeism, in which visible online activity substitutes for visible physical presence as a signal of effort. Green status indicators, fast message replies and attendance at optional meetings become performances of diligence rather than acts of work. The second is the rise of #electronic_monitoring, ranging from mild activity metrics to intrusive keystroke and screenshot tracking. Wang and colleagues found that monitoring practices act as a moderator of remote work outcomes, and the direction of that moderation is not favourable when monitoring is perceived as surveillance rather than support (Wang et al., 2021). The behavioural logic here is important for students to grasp. Monitoring signals distrust. Distrust reduces #autonomy. Reduced autonomy reduces intrinsic motivation and increases the tendency to work to the letter of the metric rather than the intent of the task. A project manager who installs surveillance software to solve a trust problem usually deepens the trust problem instead. The more constructive path found in the literature is a move towards #outcome_based_evaluation, in which what is measured is delivered value and honoured commitments rather than hours logged or presence displayed. This is easier to state than to implement, because it requires the project manager to define outcomes clearly enough to be assessed, and many project environments have historically avoided that discipline by relying on visible effort as a proxy. 2.7 Wellbeing, boundaries and the human cost The wellbeing literature is the most consistent of all the streams reviewed here, and its message is not comfortable. Kniffin and colleagues, in a broad review of workplace consequences of the pandemic, catalogued increased isolation, blurred boundaries and elevated stress alongside genuine benefits in flexibility and commuting time (Kniffin et al., 2021). Russo, Hanel and Altnickel, in a longitudinal study of software professionals, found that wellbeing and productivity were linked and that the predictors of both included factors such as autonomy, social support and the quality of the home working environment rather than remote work status alone (Russo et al., 2021). Two behavioural patterns matter most for project teams. The first is #boundary_management failure, in which the working day expands because there is no commute to end it and no building to leave. Gibbs and colleagues observed exactly this expansion of hours in their data (Gibbs et al., 2023). The second is #meeting_load. When informal coordination disappears, managers often replace it with formal meetings, and because scheduling a video call is frictionless, the number of meetings grows faster than the coordination need. The result is #video_fatigue and a working day fragmented into pieces too small for deep work. Riaz, Stankeviciute and Pinzaru, framing the post-pandemic period in terms of new work demands, connect these patterns to the wider phenomena of resignation and disengagement that organisations reported in the years after 2021 (Riaz et al., 2024). For project managers, the practical translation is blunt: a schedule that assumes eight productive hours per person per day is a fiction in most hybrid environments, and a plan built on that fiction will slip. 2.8 Competency shifts in the project manager role Finally, a smaller but growing stream examines what the pandemic did to the project manager's own skill set. Danielak and Wysocki studied competence development across selected project management areas during the remote period and reported that some competences, particularly those relating to communication, digital tooling and stakeholder handling, were developed intensively under pressure, while others received less attention (Danielak and Wysocki, 2022). Zachosova, Zhyvko and Zanora, approaching the topic from a risk and security angle, emphasise the human risks introduced when project teams operate remotely, including risks to information security and to organisational cohesion (Zachosova et al., 2021). The practitioner oriented literature, including Neeley's widely read treatment of remote work leadership, converges on a similar list: deliberate communication, explicit norm setting, trust building through small and frequent commitments, and attention to inclusion (Neeley, 2021). Kane and colleagues, writing on the redesign of the post-pandemic workplace, similarly argue that the design task facing managers is not a technology task but an organisational one (Kane et al., 2021). 2.9 Summary of the gap Across these streams, three observations stand out. The evidence is strong that behaviour changed. The evidence is weaker on which changes are durable and which were artefacts of crisis conditions. And the evidence is weakest of all on what project managers should actually do differently, expressed in terms concrete enough to teach. The framework proposed in the next section is an attempt to move towards that third level. 3. Theoretical Foundations and Conceptual Framework 3.1 Four supporting theories Media richness and media synchronicity. The classical position is that communication media differ in their capacity to carry cues, provide immediate feedback and support natural language, and that ambiguous tasks require richer media. Post-pandemic evidence complicates this. Video is theoretically rich but practically depleted, because it strips out the peripheral cues that make face to face interaction efficient while adding cognitive load. Meanwhile, written asynchronous channels, which the classical theory treats as lean, turn out to have compensating strengths: they create a record, they allow reflection, and they lower the cost of participation for people who are less confident speaking in groups. The theory therefore needs to be applied with more care than simply ranking channels from rich to lean. Work design theory. Wang and colleagues demonstrate the value of treating remote work outcomes as functions of job characteristics rather than of location alone (Wang et al., 2021). Autonomy, social support, feedback, workload and monitoring are design levers. A project manager who controls these levers can produce good outcomes in a distributed team and bad outcomes in a co-located one. This framing is empowering, because it moves the discussion from complaint to design. The #job_demands_resources logic that sits behind it also explains why the same amount of remote work can energise one team member and exhaust another. Coordination theory. Coordination is the management of dependencies between activities. In a co-located project, many dependencies are managed implicitly through shared awareness. Distance destroys that shared awareness and forces dependencies to be managed explicitly, through documentation, tickets, boards and scheduled synchronisation. De Souza Santos and Ralph's grounded theory captures this transition well (de Souza Santos and Ralph, 2022). The key insight is that explicit coordination is more expensive per unit but more scalable and more auditable, which is why distributed teams often end up with better documentation than co-located ones. Social capital and trust theory. Social capital is the value embedded in relationships. Yang and colleagues show that distributed work erodes the bridging component of social capital, the connections across groups, faster than the bonding component within groups (Yang et al., 2022). Because project management is fundamentally an integrative discipline that lives at the boundaries between specialisms, the erosion of bridging #social_capital is arguably the single most serious structural threat identified in this literature. 3.2 The Behavioural Shift Framework Drawing these together, this article proposes a framework of six shifts. Each shift is stated as a movement from a pre-pandemic default behaviour to a post-pandemic default behaviour, with an associated risk and an associated opportunity. Shift From To Main risk Main opportunity 1. Coordination mode Ambient and spontaneous Written, explicit, asynchronous Slower resolution of ambiguity Searchable, auditable project record 2. Network shape Broad and porous Deep and siloed Cross workstream risks go unseen Stronger within team focus 3. Trust basis Relational and presence based Task based and evidence based New joiners struggle to build standing Fairer assessment of quiet contributors 4. Control mode Presence and effort Outcomes and commitments Surveillance creep and gaming of metrics Clearer definitions of done and value 5. Knowledge location Human memory and corridors Repositories and systems Documentation debt and information overload Reduced key person dependency 6. Boundary condition Physical separation of work and home Temporal and self imposed separation Overwork, fatigue and burnout Genuine flexibility and improved retention Two properties of this framework are worth stating. First, the shifts are interdependent. A failure to make Shift 4 towards outcome based control usually forces Shift 6 to fail as well, because people who are evaluated on presence will simulate presence at the cost of their own boundaries. Second, none of the shifts is automatically good or bad. Each carries both risk and opportunity, and which one dominates depends on deliberate managerial design. This is the central practical claim of the article. 4. Methodology 4.1 Research design This study uses an integrative literature review design. An integrative review is appropriate when a field is fragmented across disciplines, when both empirical and conceptual work is relevant, and when the goal is to generate a synthesising framework rather than to estimate an effect size. A strict meta analysis was not possible here, because the underlying studies use incompatible designs, from randomised trials to grounded theory to telemetry analysis, and incompatible outcome measures. The review follows a structured, transparent and reproducible procedure, but it does not claim to be an exhaustive systematic review in the clinical sense. 4.2 Search strategy Literature was identified from major indexed databases covering management, information systems, psychology and computing, together with targeted searching of leading journals in project management and organisational behaviour. Search strings combined three concept blocks: Block A (context): post-pandemic, COVID-19, remote work, working from home, hybrid work, distributed work, telework Block B (unit): project management, project team, virtual team, agile team, software team, team administration Block C (behaviour): behaviour, collaboration, communication, coordination, trust, psychological safety, monitoring, wellbeing, leadership Backward and forward citation tracing from a small set of highly cited anchor studies was then used to capture work that the keyword search missed. This is an important step, because terminology in this area is unstable, and different disciplines describe the same phenomenon in different words. 4.3 Inclusion and exclusion criteria Included: peer reviewed journal articles, peer reviewed conference papers from recognised venues, scholarly books and authoritative professional standards; published between 2021 and 2025; written in English; concerned with knowledge work, project work or team work; reporting empirical findings or presenting a substantive conceptual contribution. Excluded: studies published before 2021, on the grounds that findings from the acute emergency period of 2020 are not reliably transferable to the current voluntary and planned hybrid environment; opinion pieces and vendor white papers without a research base; studies of purely physical or frontline work that cannot be performed remotely; studies that used remote work only as a control variable rather than as a subject of enquiry. 4.4 Screening and quality appraisal Records were screened first by title and abstract and then by full text. Quality appraisal considered the clarity of the research question, the appropriateness of the design, the transparency of the data, the plausibility of the causal claims and the presence of limitations. Studies with strong causal identification, such as randomised trials and natural experiments, were given more evidential weight than cross sectional surveys with self reported outcomes. Qualitative studies were weighted according to the richness of their data and the transparency of their coding, not according to sample size. 4.5 Synthesis approach Thematic synthesis was used. Findings were extracted into a structured matrix recording the study context, method, sample, main findings and reported mechanisms. Initial codes were generated inductively from the findings, then grouped into candidate themes, then tested against the full evidence set to check that each theme was supported by more than one study and more than one method. Themes that survived this test became the six shifts presented in Section 3.2 and elaborated in Section 5. 4.6 Reflexivity and limitations of the method Two biases deserve acknowledgement. First, the published literature over samples technology and knowledge intensive sectors, particularly software engineering, because these sectors are easy to instrument and were early adopters of remote work. Construction, manufacturing, healthcare and public infrastructure projects are less well represented, and the conclusions should be transferred to those contexts with care. Second, publication bias is likely to favour studies that find effects over studies that find nothing. Where the review reports an absence of effect, it relies on studies that explicitly tested for equivalence, such as the null equivalence tests reported by Bloom and colleagues (Bloom et al., 2024). 5. Findings 5.1 Shift one: from ambient coordination to explicit written coordination The most consistently reported behavioural change is that project coordination moved from talk to text. In a co-located project, a very large share of coordination is ambient. It happens because people can hear each other, see each other's screens, notice each other's presence and interrupt each other cheaply. None of this coordination is planned, recorded or costed, and because it is invisible, its value is systematically underestimated until it disappears. When it disappeared, teams responded in one of two ways. Some teams tried to recreate it synchronously, by scheduling more meetings and by keeping video calls open for long periods. Others rebuilt it in writing, by moving decisions into tickets, by keeping decision logs, by writing longer commit messages and by adopting a norm that if something is not written down it did not happen. The evidence suggests that the second strategy is more sustainable, although it is more costly at the beginning. Yang and colleagues quantified the direction of travel, showing a measurable increase in asynchronous communication and a decrease in synchronous communication after the transition to firm wide remote work (Yang et al., 2022). De Souza Santos and Ralph documented the qualitative mechanism, describing how teams built explicit coordination structures to replace lost implicit awareness (de Souza Santos and Ralph, 2022). Three consequences follow for project administration. Documentation becomes infrastructure, not overhead. In distributed projects, a good #documentation_culture is not bureaucratic burden. It is the substitute for the office. Teams that treat the ticket, the decision record and the shared specification as a #single_source_of_truth outperform teams that treat them as paperwork to be completed after the real work is done. Ambiguity resolution slows down. A question that took thirty seconds to answer across a desk may take a day to answer across a chat platform, because the answerer is in a meeting, in a different time zone or in focused work. Project schedules that were calibrated on office response times are therefore systematically optimistic. Buffer must be added not for the work itself, but for the waiting between pieces of work. Written norms become a managerial deliverable. #communication_norms that were previously implicit must be written and enforced. Which channel is used for which purpose. What response time is expected on each channel. What counts as an emergency. Whether cameras are expected. What #core_hours exist, if any. Teams that leave these questions unanswered do not thereby achieve freedom. They achieve anxiety, because each member invents a private standard and then judges colleagues against it. 5.2 Shift two: from porous networks to siloed networks The second finding is structural rather than individual. Distributed work reshapes who talks to whom. Yang and colleagues found that collaboration networks became more static and more siloed, with fewer newly formed connections and a greater concentration of communication within existing clusters (Yang et al., 2022). This is a rational response to cost. Maintaining a strong tie is expensive but valuable. Maintaining a weak tie is cheap in an office and expensive online, so when the cost rises, weak ties are the first to be cut. For projects, this creates a specific and predictable failure mode. Interfaces between workstreams are the places where projects most often fail, because they are the places where responsibility is ambiguous. Interfaces are also, precisely, weak tie territory. The tester and the infrastructure engineer who solved a subtle integration problem last quarter did so because they happened to be in the same room. If they never speak now, the same problem will be solved twice, or not at all. The evidence also suggests that #onboarding is the sharpest edge of this problem. New joiners have no existing ties to preserve. They arrive into a network that has stopped forming new connections and are expected to build a professional identity through a rectangle on a screen. Agren and colleagues noted the erosion of the informal social and learning practices on which newcomers most rely (Agren et al., 2022). The transfer of #tacit_knowledge, which historically happened through observation and proximity, has no natural substitute in a fully distributed environment and must therefore be manufactured deliberately. Practically, this means the project manager acquires a new duty that did not previously exist as a distinct activity: the deliberate engineering of connection. Rotating pairings, cross workstream reviews, explicit introductions, shared demonstration sessions and periodic in person gatherings are not social luxuries. They are network maintenance, and they should appear in the plan and in the budget. 5.3 Shift three: from relational trust to evidence based trust The third shift concerns how trust is now built. In an office, trust accumulates passively. You see that a colleague arrives, concentrates, helps others and behaves consistently, and over months you form a judgement without ever consciously assessing anyone. In a distributed team, that passive channel is closed. Trust must be built actively, and it is built out of different material: kept commitments, visible work, responsive communication and reliable quality. This is not a straightforward loss. It has a real equity benefit. Presence based trust rewards the people who are visible, extroverted, physically present and socially fluent, which is not the same set as the people who do good work. Evidence based trust rewards delivery. Bloom and colleagues found that the retention benefits of hybrid work were strongest precisely among groups that are often disadvantaged by presence based cultures, including non managers and employees with long commutes (Bloom et al., 2024). But there are two serious risks. The first is #proximity_bias. In hybrid arrangements, where some team members are in the office and some are not, informal information and informal influence tend to concentrate among those physically near the decision maker. The person in the room hears the aside. The person on the screen does not. Over time this produces unequal access to opportunity that is invisible to the manager producing it, because the manager experiences it as convenience rather than as favouritism. #hybrid_meeting_equity is the operational response: if one participant is remote, the meeting is run as a remote meeting, with all contributions flowing through the same channel. The second risk is that evidence based trust degrades into metric based trust. If trust depends on visible evidence, and the only cheap evidence is activity data, then trust becomes a function of dashboards. This is how well intentioned #accountability turns into surveillance. The literature is consistent that intrusive monitoring undermines the very autonomy that makes remote work valuable (Wang et al., 2021). The synthesis is therefore that #virtual_trust in the post-pandemic project is best built through small, frequent, visible and honoured commitments rather than through either social bonding alone or measurement alone. #swift_trust gets a new team started. Delivery keeps it alive. 5.4 Shift four: from presence based control to outcome based control The fourth shift is the one that project management as a discipline is best equipped to handle, and yet is often handled worst. Traditional supervision uses presence as a proxy for effort and effort as a proxy for contribution. Both proxies were always weak, but they were cheap. Remove physical co-location and the first proxy becomes unavailable, at which point organisations face a choice: rebuild a digital version of presence, or move to genuine outcome measurement. Many organisations chose to rebuild presence, through activity indicators, availability status and mandatory camera policies. The behavioural result was #digital_presenteeism, which costs real hours, generates real stress and measures nothing useful. It also actively harms delivery, because a person who must appear available cannot enter the sustained concentration that complex project work requires. The alternative requires more from the project manager. #outcome_based_evaluation demands that outcomes be defined clearly enough to be observed, that they be agreed rather than imposed, and that they be assessed honestly. This is difficult. It exposes weak planning. A project manager who cannot say what "done" means for a given piece of work has been relying on presence to hide that gap. The wider governance implication is that #project_governance must shift its attention from inputs to commitments. Steering committees that ask how many hours were spent are asking a question that distributed work cannot answer meaningfully. Steering committees that ask what was promised, what was delivered, what was learned and what changed are asking questions that work equally well in every location. A necessary corollary is that #autonomy must genuinely increase. Outcome based control without autonomy is simply the same supervision with an added measurement burden. The bargain has to be real: the team gains discretion over how and when the work is done, and in return accepts clear responsibility for what is delivered. 5.5 Shift five: from human memory to system memory The fifth shift is that project knowledge has migrated out of people's heads and out of corridor conversations, and into systems. In practice this means that the shared repository, the ticket system, the recorded decision and the searchable channel archive now carry a much greater share of the project's working memory than they did before 2020. This has three effects worth separating. It reduces key person dependency. When knowledge lives in a system, the departure of a senior engineer is less catastrophic. Zachosova and colleagues, considering the human risks of remote project teams, note that dependence on individuals is one of the principal exposures organisations face (Zachosova et al., 2021), and system memory is a partial hedge against it. It creates documentation debt. Repositories decay. A wiki page that is six months out of date is worse than no wiki page, because it is confidently wrong. Distributed teams therefore acquire a maintenance obligation that co-located teams could ignore, and this obligation needs to be planned and resourced rather than assumed. It changes the search problem. When everything is written, the difficulty moves from remembering to finding. Teams report drowning in information while simultaneously being unable to locate the specific decision they need. This is where #artificial_intelligence and #automation are beginning to matter most in practice: not as replacements for project managers, but as retrieval and summarisation layers over an information base that has grown beyond human scanning capacity. The evidence base on the effectiveness of these tools in project settings is still thin, and students should treat vendor claims with appropriate scepticism. Two further consequences follow. #data_governance becomes a project management concern rather than purely a technical one, because the project's memory is now a data asset with retention, access and quality requirements. And #cybersecurity moves from the periphery to the centre, because a project whose knowledge lives in distributed systems accessed from home networks has a materially different threat surface than one whose knowledge lives in a locked office. 5.6 Shift six: from spatial boundaries to temporal and self imposed boundaries The sixth shift is the human cost, and it is the one most likely to be underestimated in project plans. The commute, whatever its other faults, performed a psychological function. It marked a transition. Remove it, and the boundary between work and life must be maintained by the individual, using willpower, in the same physical space where the work happens. Many people cannot do this reliably, and the evidence shows the predictable result: the working day expands. Gibbs and colleagues found precisely this pattern of longer hours without proportionate output gains (Gibbs et al., 2023). Compounding this is #meeting_load. Because scheduling is frictionless, and because managers who cannot see their teams feel anxious, meetings multiply. The calendar fills with thirty minute blocks that fragment the day into intervals too short for demanding work. #video_fatigue is a real phenomenon and not merely a complaint: sustained self-view, reduced mobility, intense eye contact at unnatural distances and constant cognitive effort to interpret degraded social cues all impose measurable load. The consequences are not only individual. #burnout is a project risk. It degrades quality, increases error rates, damages relationships and eventually produces the departures that destroy schedules. Riaz and colleagues connect these new work demands directly to the disengagement and attrition phenomena of the post-pandemic labour market (Riaz et al., 2024). Kniffin and colleagues catalogue the isolation and boundary problems alongside the genuine benefits (Kniffin et al., 2021). Yet the same body of evidence contains the counterweight. Bloom and colleagues found that hybrid working improved job satisfaction and cut quit rates by about a third with no measurable damage to performance (Bloom et al., 2024). #employee_wellbeing and #productivity are not in opposition here. Flexibility that is genuine, voluntary and well designed appears to improve both. Flexibility that is nominal, that is granted while presence is still rewarded and while meetings still consume the day, produces the worst of both worlds. The distinguishing variable, once again, is design. Protected focus blocks, meeting free days, explicit norms about out of hours contact, realistic capacity assumptions in the schedule and honest acknowledgement of #time_zones in #global_virtual_teams are not perks. They are the mechanisms that make Shift Six land on the opportunity side rather than the risk side. 5.7 Cross cutting finding: the project manager's role has changed shape Across all six shifts, one further finding emerges that belongs to none of them individually. The project manager's job has moved further away from administration and further towards design. The administrative components, status collection, progress reporting, schedule updating, have been substantially absorbed by tooling and increasingly by #automation. What has grown instead is the work of designing the conditions in which distributed people can collaborate: setting norms, engineering connection, defining outcomes, protecting attention, maintaining trust and noticing the quiet signals of trouble in an environment that no longer broadcasts them. Danielak and Wysocki found that competences relating to communication, digital tooling and stakeholder handling developed intensively during the remote period, which is consistent with this reading (Danielak and Wysocki, 2022). The competency profile that follows is demanding: #e_leadership and #distributed_leadership capability, strong written communication, facilitation skill, #emotional_intelligence sufficient to detect distress through a low bandwidth channel, comfort with #digital_maturity assessment and tooling, and the analytical discipline to run #performance_measurement without sliding into surveillance. 6. Discussion 6.1 Interpreting the contradictions Students reading this literature for the first time will notice that the studies appear to disagree. Some find productivity gains, others find losses. Some find improved wellbeing, others find burnout. Some find that agile ceremonies survived, others find that agile culture eroded. These disagreements are less severe than they look, and they resolve along three lines. The task type line. Independent, well defined, individually owned work travels well across distance. Interdependent, ambiguous, negotiated work travels badly. Choudhury and colleagues studied a setting with substantial individual autonomy and found gains (Choudhury et al., 2021). Yang and colleagues studied network level collaboration and found erosion (Yang et al., 2022). Both can be true simultaneously, because they are measuring different layers of the same organisation. The volition line. Forced remote work during a pandemic, with closed schools and elevated anxiety, is a different phenomenon from chosen hybrid work in a stable period. Smite and colleagues make this distinction the centre of their argument (Smite, Moe, Klotins and Gonzalez-Huerta, 2023). Findings from 2020 describe a crisis. Findings from 2023 onwards describe a working arrangement. The design line. This is the most important. Almost every negative finding in the literature is contingent on a design choice: excessive monitoring, absent norms, unrealistic capacity assumptions, no deliberate connection building, no onboarding redesign. Almost every positive finding is contingent on the opposite. Wang and colleagues' work design framing makes this explicit (Wang et al., 2021), and it is the strongest reason to treat "does remote work work?" as a badly posed question. 6.2 What is genuinely new for project management Three things in this evidence base are not simply old virtual team findings restated. First, the scale changes the mechanism. Pre-pandemic virtual team research studied minorities of dispersed workers embedded in mostly co-located organisations, where the physical hub still carried the informal load. When the majority is distributed, there is no hub to carry it, and the informal layer does not merely thin. It disappears and has to be rebuilt from nothing. Second, hybrid is harder than fully remote. This is one of the more counter-intuitive lessons of the period. A fully distributed team has one set of rules and one experience of the meeting. A hybrid team has two, and the two are not equal. #proximity_bias, #hybrid_meeting_equity and the unequal distribution of informal information are hybrid problems, not remote problems, and they arise precisely because the arrangement looks like a compromise but behaves like a hierarchy. Third, the artefact has become the team's shared mind. In a co-located team, the board on the wall is a representation of the work. In a distributed team, the board in the system is the work, in the sense that anything not on it is effectively invisible. This elevates project tooling from a support function to a constitutive one, and it means that decisions about tooling, taxonomy, workflow states and naming conventions are decisions about how the team thinks. 6.3 The unresolved tension The literature leaves one tension genuinely unresolved, and honesty requires stating it. Individual preference and organisational coordination need do not point in the same direction. Individuals, on the whole, want more flexibility, and the evidence that this improves their satisfaction and retention is strong (Bloom et al., 2024; Barrero et al., 2021). Organisations, on the whole, want more informal cross boundary collaboration, and the evidence that distance harms this is also strong (Yang et al., 2022). Both bodies of evidence are credible. They are simply about different things. The current managerial response, blanket return to office mandates, satisfies the organisational concern by overriding the individual one, and Smite and colleagues document the resulting friction (Smite, Moe, Hildrum, Gonzalez-Huerta and Mendez, 2023). The alternative response, unconstrained individual choice, satisfies the individual concern while allowing coordination to decay quietly. Neither is a solution. The most promising direction in the literature, and it remains under-tested, is coordinated flexibility: the team, not the individual and not the executive, decides its own pattern, with the constraint that collaboration intensive activities are deliberately co-located in time even if not in space, and that periodic physical gathering is planned as a network building investment rather than as a compliance requirement. This is a hypothesis, not an established finding, and Section 9 treats it as such. 7. Implications 7.1 For project practice Write the norms down. Every distributed project should have an explicit working agreement covering channels, response expectations, meeting rules, core overlap hours, camera expectations and escalation paths. This document costs an afternoon and prevents months of friction. Plan connection as work. Allocate budget and calendar time to cross workstream contact, onboarding buddies, periodic gatherings and rotating pairings. Treat these as risk mitigation for the siloed network problem, and record them in the plan. Define done before you distribute. Outcome based control is impossible without clear definitions of completion and acceptance. If a project manager cannot articulate what finished looks like, that gap must be closed before it is exported into a distributed environment where presence can no longer disguise it. Protect attention. Cap meeting hours. Establish meeting free blocks. Require an agenda and a decision statement for every meeting. Default to asynchronous unless the topic requires negotiation, conflict resolution or creative divergence, which are the three categories that genuinely need synchronous richness. Measure delivery, not activity. Resist the temptation to install surveillance tooling. It buys a feeling of control and sells the team's motivation to pay for it. Redesign onboarding completely. Assume that nothing a new joiner needs will be absorbed by osmosis. Everything must be explicit, scheduled and paired. Rebalance the schedule. Add buffer for asynchronous latency. Reduce assumed daily productive hours to something honest. A plan that assumes the office is a plan for a place that no longer exists. 7.2 For project management education Project management curricula have historically taught tools, techniques and process areas. The evidence in this review implies that the curriculum needs to add a substantial behavioural and design component. Specifically, students should be taught: how to write clearly enough that a specification can survive without its author present; how to facilitate an inclusive hybrid meeting; how to design a working agreement; how to build trust in a team they have never physically met; how to detect distress through a low bandwidth channel; how to construct outcome definitions that are honest; and how to think about the ethics of monitoring. #training_and_development in these areas is currently uneven, and #project_manager_competencies frameworks are still catching up with what the job now requires. Assessment should also change. A student who can produce a Gantt chart but cannot run a productive asynchronous decision thread is not prepared for the work as it now exists. 7.3 For organisational policy Organisations should recognise that a location policy is a coordination policy, not a real estate policy. Smite and colleagues show the wide variation in company approaches and the tension these create (Smite, Moe, Hildrum, Gonzalez-Huerta and Mendez, 2023). Policies imposed without consultation generate resentment out of proportion to their content, because they are read as statements about trust. Three policy recommendations follow from the evidence. Grant flexibility genuinely rather than nominally, because half granted flexibility combined with presence based reward produces the worst outcomes. Set the policy at team level where possible, because coordination needs differ enormously between a research team and a delivery team. And invest the money saved on office space in the things that distributed work actually requires: better tooling, better documentation practices, periodic gatherings and manager training. Finally, #change_management applies to the shift itself. Moving a project organisation from presence based to outcome based operation is an organisational change, and it fails for the usual reasons that organisational changes fail: unclear rationale, insufficient support, and leaders who exempt themselves from the new behaviours. 8. Limitations This review has four significant limitations that should temper any confident reading of its conclusions. Sectoral bias. The evidence base is dominated by software, technology and professional services. These sectors are highly digitisable, well instrumented and culturally receptive to remote work. Construction projects, hospital transformation programmes, defence procurement and infrastructure delivery are much less well represented in the literature, and the six shifts should be treated as hypotheses rather than findings in those contexts. Geographic bias. Most of the strong causal work comes from the United States, Western Europe and China. Findings from these settings may not transfer to regions with different labour market structures, different housing conditions, different infrastructure quality and different cultural expectations regarding authority, hierarchy and #stakeholder_engagement. Temporal instability. The subject is moving. Behaviour recorded in 2021 was shaped by crisis conditions. Behaviour recorded in 2024 was shaped by a labour market with different power dynamics. Behaviour in 2026 is being shaped by the rapid introduction of generative tooling into everyday collaboration, a development too recent to have a mature research literature. Method limitations. This is an integrative review conducted by a single synthesis process, without independent double screening or formal inter-rater reliability testing. It cannot exclude selection bias in what was included, and it makes no claim to exhaustive coverage. It also inherits the publication bias of the underlying literature. 9. Future Research Agenda Six directions follow directly from the gaps identified above. 1. Longitudinal studies of network erosion. Yang and colleagues gave us a snapshot of siloing during the transition (Yang et al., 2022). What is now needed is a multi year study of whether networks recover, stabilise or continue to narrow, and whether hybrid arrangements repair bridging ties or merely slow their decay. 2. Non-digital sectors. Rigorous studies of hybrid project administration in construction, healthcare and public infrastructure are largely absent. These sectors have physical work that cannot be relocated but administrative and coordination work that can, and this mixed condition is theoretically interesting and practically urgent. 3. Onboarding and tacit knowledge transfer. This is the sharpest practical problem and the thinnest evidence base. Controlled comparisons of onboarding designs in distributed project teams would have immediate practical value. 4. Coordinated flexibility. The hypothesis that team level scheduling coordination outperforms both individual choice and executive mandate is plausible, widely advocated and largely untested. It is well suited to field experimentation. 5. Monitoring, autonomy and performance. The relationship between #electronic_monitoring intensity, perceived autonomy, trust and delivery outcomes deserves careful causal work rather than the cross sectional correlations that currently dominate. 6. Generative tools in project coordination. As #artificial_intelligence assistants are embedded into meeting platforms, ticket systems and document repositories, they will change what project managers do and what teams believe. Early, careful, sceptical empirical work here would be valuable, particularly on questions of accuracy, over-reliance, accountability and the possible further reduction of human to human contact. #digital_transformation of the project office is proceeding much faster than the evidence about its effects. An additional methodological plea: the field would benefit enormously from more studies that combine behavioural telemetry with qualitative interpretation. Telemetry tells us what changed. Interviews tell us why. Very few studies do both, and the ones that do are among the most useful in this review. 10. Conclusion The pandemic did not merely move project work from the office to the home. It removed a support structure that the project management discipline had relied on without ever naming it: the ambient, unplanned, unmeasured human interaction that a shared physical space produces for free. The evidence reviewed here suggests that project teams have adapted, but that they have adapted by changing their behaviour, not merely their location. Coordination has moved from talk to text. Networks have become deeper and narrower. Trust has shifted from presence to evidence. Control has shifted, unevenly and incompletely, from effort to outcome. Project memory has moved from human heads to digital systems. And the boundary that protected personal life has become something each individual must defend alone. None of these shifts is inherently good or bad. Each carries a risk and an opportunity, and which one materialises depends on a set of design choices that sit squarely within the project manager's authority. This is the central and, for students, the hopeful conclusion of the review. The literature does not show that distributed project work is doomed, nor that it is automatically superior. It shows that it is far more sensitive to deliberate design than co-located work ever was, because the office used to compensate quietly for a great deal of poor management, and it no longer can. The project manager who understands this stops asking whether the team should be in the office, and starts asking a better question: what behaviours does this project require, and what conditions must I create so that those behaviours happen? That question is harder, and it is the right one. #project_success in the post-pandemic era will belong to the managers who design for behaviour, not to those who legislate for attendance. The discipline's next decade of #resilience and #future_of_work capability depends on taking that seriously. References Agren, P., Knoph, E. and Berntsson Svensson, R. (2022) Agile software development one year into the COVID-19 pandemic. Empirical Software Engineering, 27(6), article 121. Barrero, J. M., Bloom, N. and Davis, S. J. (2021) Why working from home will stick. National Bureau of Economic Research Working Paper No. 28731. https://doi.org/10.3386/w28731 Bick, A., Blandin, A. and Mertens, K. (2023) Work from home before and after the COVID-19 outbreak. American Economic Journal: Macroeconomics, 15(4), 1-39. Bloom, N., Han, R. and Liang, J. (2024) Hybrid working from home improves retention without damaging performance. Nature, 630(8018), 920-925. https://doi.org/10.1038/s41586-024-07500-2 Choudhury, P., Foroughi, C. and Larson, B. (2021) Work-from-anywhere: The productivity effects of geographic flexibility. Strategic Management Journal, 42(4), 655-683. https://doi.org/10.1002/smj.3251 Conboy, K., Moe, N. B., Stray, V. and Gundelsby, J. H. (2023) The future of hybrid software development: Challenging current assumptions. IEEE Software, 40(2), 26-33. Danielak, W. and Wysocki, R. (2022) The impact of remote work during the COVID-19 pandemic on the development of competences in selected areas of project management. Annales Universitatis Mariae Curie-Sklodowska, sectio H, Oeconomia, 56(2), 7-20. https://doi.org/10.17951/h.2022.56.2.7-20 De Souza Santos, R. E. and Ralph, P. (2022) A grounded theory of coordination in remote-first and hybrid software teams. In: Proceedings of the 44th International Conference on Software Engineering (ICSE 2022), pp. 25-35. Ford, D., Storey, M. A., Zimmermann, T., Bird, C., Jaffe, S., Maddila, C., Butler, J. L., Houck, B. and Nagappan, N. (2022) A tale of two cities: Software developers working from home during the COVID-19 pandemic. ACM Transactions on Software Engineering and Methodology, 31(2), 1-37. Gibbs, M., Mengel, F. and Siemroth, C. (2023) Work from home and productivity: Evidence from personnel and analytics data on information technology professionals. Journal of Political Economy Microeconomics, 1(1), 7-41. https://doi.org/10.1086/721803 Kane, G. C., Nanda, R., Phillips, A. and Copulsky, J. (2021) Redesigning the post-pandemic workplace. MIT Sloan Management Review, 62(3), 12-14. Kniffin, K. M., Narayanan, J., Anseel, F., Antonakis, J., Ashford, S. P., Bakker, A. B. and colleagues (2021) COVID-19 and the workplace: Implications, issues, and insights for future research and action. American Psychologist, 76(1), 63-77. https://doi.org/10.1037/amp0000716 Neeley, T. (2021) Remote Work Revolution: Succeeding from Anywhere. New York: Harper Business. Project Management Institute (2021) A Guide to the Project Management Body of Knowledge (PMBOK Guide), Seventh Edition, and The Standard for Project Management. Newtown Square, PA: Project Management Institute. Riaz, Z., Stankeviciute, Z. and Pinzaru, F. (2024) Editorial: New work demands and managing employee well-being in the post-pandemic world. Frontiers in Psychology, 15, article 1392687. https://doi.org/10.3389/fpsyg.2024.1392687 Russo, D., Hanel, P. H. P. and Altnickel, S. (2021) Predictors of well-being and productivity among software professionals during the COVID-19 pandemic: A longitudinal study. Empirical Software Engineering, 26(4), article 62. https://doi.org/10.1007/s10664-021-09945-9 Smite, D., Moe, N. B., Hildrum, J., Gonzalez-Huerta, J. and Mendez, D. (2023) Work-from-home is here to stay: Call for flexibility in post-pandemic work policies. Journal of Systems and Software, 195, article 111552. https://doi.org/10.1016/j.jss.2022.111552 Smite, D., Moe, N. B., Klotins, E. and Gonzalez-Huerta, J. (2023) From forced working-from-home to voluntary working-from-anywhere: Two revolutions in telework. Journal of Systems and Software, 195, article 111509. https://doi.org/10.1016/j.jss.2022.111509 Vartiainen, M. (2021) Telework and remote work. In: Oxford Research Encyclopedia of Psychology. Oxford: Oxford University Press. Wang, B., Liu, Y., Qian, J. and Parker, S. K. (2021) Achieving effective remote working during the COVID-19 pandemic: A work design perspective. Applied Psychology: An International Review, 70(1), 16-59. https://doi.org/10.1111/apps.12290 Yang, L., Holtz, D., Jaffe, S., Suri, S., Sinha, S., Weston, J., Joyce, C., Shah, N., Sherman, K., Hecht, B. and Teevan, J. (2022) The effects of remote work on collaboration among information workers. Nature Human Behaviour, 6(1), 43-54. https://doi.org/10.1038/s41562-021-01196-4 Zachosova, N., Zhyvko, Z. and Zanora, V. (2021) Human risks of project management involving remote teams in the context of the enterprise economic and financial security. Financial and Credit Activity: Problems of Theory and Practice, 4(35), 514-521. https://doi.org/10.18371/fcaptp.v4i35.222530 Hashtags #PostPandemicProjectManagement #RemoteTeamAdministration #DigitalCollaboration #HybridProjectTeams #VirtualTeamManagement #ProjectManagementResearch #RemoteWorkBehaviour #DistributedTeams #CollaborationTechnology #WorkplaceBehaviouralChange #AgileRemoteTeams #ProjectLeadership #FutureOfProjectWork #AcademicResearchArticle #ScopusStyleReview

Public and private administrations are being asked to do two things at once. They are asked to hold a steady course toward long-term strategic goals, and they are asked to absorb a new class of technology that changes every few months. The newest form of this technology is agentic artificial intelligence: software that does not simply answer a question but plans a sequence of actions, calls other tools, and works toward a goal with limited human supervision. This article examines the tension between strategic continuity and rapid technological absorption, and asks how administrations can move quickly without losing direction. The study is an integrative conceptual review. It synthesises literature from public administration, strategic management, and information systems published mainly between 2021 and 2025, and it reads that literature through three theoretical lenses: dynamic capabilities, organisational ambidexterity, and institutional theory. From this synthesis, the article identifies six recurring pressure points where speed and strategy collide: the accountability gap created by autonomous action, the readiness of data infrastructure, the reshaping of administrative discretion, workforce capacity and change fatigue, procurement and dependency on vendors, and the maintenance of public legitimacy. The article then proposes a framework, the Anchored Agility Framework, built on five pillars: strategic anchoring, modular adoption, layered guardrails, institutionalised learning loops, and legitimacy by design. A five-stage maturity model translates the framework into a practical sequence for administrators. The central argument is that agility is not the opposite of strategy. Agility is the capacity to change the means of implementation quickly while the ends remain stable. Administrations that treat every new model release as a strategic event will exhaust themselves. Administrations that ignore the shift will fall behind. The path between these errors requires deliberate design, not improvisation. Keywords: administrative agility; agentic artificial intelligence; strategic management; dynamic capabilities; public sector innovation; algorithmic governance 1. Introduction For most of the last century, the strength of a large administration was its steadiness. Rules were written down, roles were fixed, and decisions followed a known path. This steadiness was not an accident. It was the point. It made outcomes predictable, protected citizens and employees from arbitrary treatment, and allowed complex work to continue even when individual people left. Steadiness, however, has a cost. It makes the organisation slow to change, and the world does not always wait. The arrival of #agentic_AI has made this old problem urgent again. Earlier waves of digital technology mostly automated tasks that humans had already defined. A payroll system calculated what a clerk would have calculated. A case management database stored what a file cabinet would have stored. Agentic systems are different in kind, not only in degree. They accept a goal, break it into steps, decide which tools to use, act, observe the result, and adjust. In other words, they exercise a limited form of #autonomy inside the administrative process itself. This creates a genuine dilemma for anyone responsible for a long-term plan. Strategic documents in #public_administration are typically written on horizons of three, five, or ten years. They are approved by boards, councils, or legislatures. They are tied to budgets that cannot easily be moved. Meanwhile, the underlying capability of AI systems has been changing on a cycle measured in months. A strategy written with confidence in one year may rest on assumptions about what machines can and cannot do that are simply false eighteen months later. The instinctive responses to this dilemma are both wrong. The first response is to freeze. Administrators say that they will wait until the technology settles, until standards are clear, until the risks are understood. This sounds prudent. In practice it means the organisation loses the internal knowledge it would need to evaluate the technology at all, and becomes dependent on whichever supplier arrives with the loudest promise. The second response is to chase. Administrators sign agreements for every promising tool, launch dozens of pilots, and announce transformation. This produces motion without progress: a scattering of disconnected systems, none of which is properly evaluated, and a workforce that no longer believes any announced change is real. This article argues for a third position, which it calls administrative agility. Administrative agility is the organised capacity to change how goals are pursued at high speed while keeping what is being pursued stable and publicly defensible. It is not the absence of planning. It is a particular kind of planning, one that separates durable commitments from disposable methods, and that builds the internal machinery to make the disposable parts genuinely disposable. #strategic_alignment, in this reading, is not a document. It is a repeated act of connecting fast-moving technical choices back to slow-moving public commitments. The article makes four contributions. First, it clarifies what is actually new about agentic systems, separating substance from marketing language, because much of the current confusion in #digital_government stems from treating all AI as one thing. Second, it maps the specific points at which the speed of agentic adoption collides with the requirements of strategic and democratic accountability. Third, it develops the Anchored Agility Framework as an organising structure for administrators facing these collisions. Fourth, it offers a maturity model that turns the framework into a sequence of concrete steps, along with an agenda for future empirical work. The remainder of the article proceeds as follows. Section 2 sets out the conceptual background of agility and strategy in administrative settings. Section 3 defines agentic AI with precision. Section 4 develops the theoretical framework. Section 5 explains the review method. Sections 6 and 7 present the core analysis of the speed and stability tension and the six pressure points. Sections 8 and 9 present the framework and maturity model. Section 10 discusses documented patterns from practice. Sections 11 to 13 discuss implications, limitations, and conclusions. 2. Conceptual Background: Agility and Strategy in Administrative Organisations 2.1 The origins of the agility concept The idea of #organizational_agility entered management thinking through manufacturing, where it described the ability to switch production quickly in response to changing demand. It was later broadened to describe a firm's ability to sense changes in its environment and respond to them faster than competitors. In the information systems literature, agility became closely linked to the configuration of digital infrastructure: organisations with modular, well-documented, loosely coupled systems could reconfigure themselves; organisations with tangled, undocumented, tightly coupled systems could not, regardless of how much they wanted to. Two features of this history matter for the present discussion. First, agility was always understood as a capability, not a mood. It is built through investment in specific assets, such as interoperable systems, cross-functional teams, and decision rights that are clear enough to allow fast choices. Talking about agility without building these assets produces slogans, not speed. Second, agility was always understood as bounded. No organisation is agile in every dimension. The question is always which dimensions of the organisation must be able to move quickly, and which must remain stable so that the moving parts have something to move against. 2.2 Why administrative settings are different Applying agility to administration requires care, because administrative organisations are not firms. They differ in at least four ways that shape what agility can mean. The first difference is the source of legitimacy. A firm is judged mainly by results. An administration is judged by results and by the propriety of the process that produced them. A decision that is correct but reached through an unaccountable process is a failure in administrative terms, even if the outcome is good. This means that any speed gained by removing procedural steps is not a pure gain; it is a trade against legitimacy. The second difference is the multiplicity of principals. A firm's management answers, in the end, to owners. An administration answers to elected officials, to oversight bodies, to courts, to professional norms, and to the public directly. These principals do not always want the same thing, and they do not move at the same speed. An elected official may want a visible result before an election. A court may want a decision that will survive review in five years. #bureaucracy, often criticised as slow, is partly a mechanism for holding these conflicting demands together. The third difference is the difficulty of exit. A dissatisfied customer changes supplier. A dissatisfied citizen usually cannot change the agency that issues their licence, assesses their tax, or decides their benefit claim. Because exit is unavailable, the burden on the quality and fairness of the process is much higher. Errors are not corrected by the market. They are corrected, if at all, by internal control, oversight, and complaint. The fourth difference is the structure of budgets. Public budgets are typically annual, appropriated for defined purposes, and difficult to redirect. This creates a specific form of rigidity that has nothing to do with the attitudes of staff. An agency may see clearly that a new approach is better and still be unable to fund it until the next cycle. Reforming this constraint is itself a slow, political act. 2.3 Strategy under conditions of technological uncertainty Classic #strategic_planning assumes that the planner can form a reasonable picture of the future environment. When that assumption holds, planning is powerful: it aligns resources, sets priorities, and allows commitment. When the assumption fails, planning can be actively harmful, because it locks resources into a picture of the world that no longer exists. The literature on strategy under uncertainty suggests a partial solution: distinguish between levels of commitment. Some commitments are irreversible and expensive, such as constructing a facility or signing a decade-long contract. Others are cheap and reversible, such as running an experiment or training staff in a transferable skill. Under high uncertainty, the rational move is to shift weight toward reversible commitments and to delay irreversible ones until the uncertainty resolves, while accepting that some irreversible commitments cannot be delayed. This is directly relevant to agentic AI. Many current adoption decisions are being taken as if they were reversible when they are not. A multi-year contract that embeds a single vendor's agent framework into the core of a case processing system is not a pilot. It is a structural commitment, and it will shape what the organisation can do for years. 2.4 Path dependency and the weight of legacy The concept of #path_dependency explains why administrations often cannot simply choose the best available option. Past choices constrain present ones through sunk investment, accumulated skills, established data formats, and the expectations of staff and citizens. A department that built its records around one data model in 2005 will find that this model quietly determines what analytics are possible in 2026. This is not an argument for despair. It is an argument for treating infrastructure choices as strategic in themselves. The organisations that will be able to adopt agentic systems fastest in three years are not necessarily the ones running the most pilots today. They are the ones that are cleaning their data, documenting their processes, and building interfaces between their systems. This work is unglamorous and rarely announced. It is also the actual precondition for #ambidexterity, the capacity to run the existing operation reliably while building the next one at the same time. 3. What Agentic Artificial Intelligence Actually Is 3.1 From prediction to generation to action It helps to see the current moment as the third step in a sequence. The first step was predictive machine learning. A system was trained on historical data to estimate a value or a category: the likelihood that a claim is fraudulent, the expected demand for a service, the risk score attached to a case. These systems produced numbers. A human then decided what to do with the number. The second step was generative AI. Systems based on #large_language_models produced fluent text, code, images, and summaries in response to instructions. The output was no longer a number but an artefact. This changed the range of tasks that could be supported, but the basic pattern held: a human asked, the system produced, and the human decided. The third step is the agentic pattern. Here the system is given a goal rather than a single instruction. It plans a sequence of steps, selects and calls external tools, observes the results, revises its plan, and continues until it judges the goal met or a limit reached. The characteristic features are #goal_directed_behavior, #tool_use, memory across steps, and the ability to act on the world rather than only describe it. The system can query a database, write a record, send a message, submit a form, or trigger another system. 3.2 The components of an agentic system An agentic deployment typically involves four layers, and confusing them is a common source of error in procurement and governance discussions. The reasoning layer is the model that plans and decides. It is usually a general-purpose model, not built for the organisation, and it may be replaced with a newer version without any change to the surrounding software. The tool layer is the set of actions the agent is permitted to take. This is where the real risk lives. A model that can only read is a different object, in governance terms, from a model that can write to a live register, even if the underlying reasoning model is identical. The orchestration layer manages the loop: how many steps are allowed, what happens on failure, when a human is consulted, how state is stored. #agent_orchestration is the layer where most engineering effort is actually spent, and where most organisational control can be exercised. The supervision layer determines how humans interact with the process. This ranges from full autonomy through to a strict #human_in_the_loop design in which the agent proposes and a person approves each consequential action. When organisations debate whether to "adopt agentic AI," they are usually debating the reasoning layer, which is the layer they control least and which matters least for governance. The decisions that determine risk and value are made in the tool, orchestration, and supervision layers, which the organisation controls almost entirely. 3.3 Multi-agent configurations Some deployments involve several agents that divide work between them, with one agent coordinating others or reviewing their output. #multi_agent_systems can improve performance on complex tasks, but they multiply failure modes. Errors compound across steps, agents can reinforce each other's mistakes, and the trace of who decided what becomes difficult to reconstruct. From an administrative point of view, a multi-agent configuration is not a technical refinement. It is a change in the structure of decision-making inside the organisation, and it should be reviewed as such. 3.4 What is genuinely new, and what is not Three things are genuinely new. First, the unit of delegation has shifted from the task to the objective, which means the organisation is no longer specifying the method. Second, the systems act, so errors have direct consequences rather than passing through a human filter. Third, the pace of capability change is fast enough that assumptions embedded in policy documents expire quickly. Several things are not new, despite frequent claims to the contrary. The need for data quality is not new. The need for clear accountability for administrative decisions is not new. The tendency of technology projects to fail for organisational rather than technical reasons is not new. The pattern in which vendors promise transformation and deliver a chatbot is not new. Recognising the continuity here is not cynicism. It is what allows an administration to apply hard-won institutional knowledge rather than starting from zero every time a new term becomes popular. 4. Theoretical Framework The article draws on three theoretical traditions. Each explains part of the problem, and none explains all of it. 4.1 Dynamic capabilities The #dynamic_capabilities perspective distinguishes ordinary capabilities, which allow an organisation to do what it currently does efficiently, from dynamic capabilities, which allow it to change what it does. Dynamic capabilities are usually grouped into three activities. #sensing is the capacity to detect relevant change in the environment early and accurately. In the agentic AI context, sensing is not about reading technology news. It is about maintaining enough internal technical literacy to distinguish a genuine capability shift from a marketing cycle, and enough contact with frontline operations to know which administrative processes are actually candidates for delegation. #seizing is the capacity to commit resources to an opportunity once detected. This is where many administrations fail. They sense correctly and then cannot act, because budgets are fixed, procurement takes eighteen months, and no one has authority to decide. Seizing capacity is built through pre-authorised funding envelopes for experimentation, framework contracts that allow rapid engagement of suppliers, and clear delegation of decision rights. #reconfiguring is the capacity to reshape the organisation once a new approach proves itself, which includes the painful work of shutting down what the new approach replaces. Administrations are generally much better at addition than at subtraction. A pilot that succeeds and is then run in parallel with the old process forever has not reconfigured anything; it has added cost. 4.2 Organisational ambidexterity Ambidexterity theory holds that organisations must simultaneously pursue #exploitation, the efficient running of current activities, and #exploration, the search for new ones. These two modes require different structures, incentives, time horizons, and tolerances for failure, and they compete for the same scarce resources: attention, money, and talent. The theory offers two main solutions. Structural ambidexterity separates the two modes into different units, with an innovation team protected from the operational pressures of the main organisation. This protects exploration but risks producing innovations that the main organisation never absorbs. Contextual ambidexterity asks the same people to do both, supported by a culture and set of systems that make switching possible. This improves absorption but risks exploration being crushed by the urgency of daily operations, which always wins in the short run. For agentic AI, neither pure solution works. A separated AI unit will build agents that operational staff distrust and never use. A purely contextual approach will produce nothing, because case officers with a backlog will not experiment. The practical answer, developed in Section 8, is a hybrid: a small central capability that builds shared infrastructure and guardrails, combined with distributed experimentation inside operational units under those guardrails. 4.3 Institutional theory Institutional theory explains a phenomenon that purely rational accounts cannot: organisations frequently adopt practices that do not improve performance, because adoption confers legitimacy. #institutional_isomorphism describes the process by which organisations in a field come to resemble one another, through coercive pressure from regulators and funders, mimetic copying of peers under uncertainty, and normative pressure from professions and consultants. This is not a marginal observation in the current environment. It is central. A great deal of agentic AI adoption in administration is driven by the need to be seen to be adopting. Strategies are published because peer organisations have published strategies. Pilots are launched because a minister has referred to AI in a speech. Consultants recommend the same architecture to every client. The result is a field converging on similar practices for reasons that have little to do with evidence. Institutional theory also supplies the concept of decoupling: the organisation adopts the practice formally, at the level of policy and announcement, while the operational core continues largely unchanged. Decoupling is often described as hypocrisy. It is better understood as a defence mechanism, protecting the operational core from disruption while satisfying external demands for conformity. It is also, for the purposes of this article, a diagnostic warning. An administration with an impressive AI strategy, a governance board, and no measurable change in any service process is not failing to implement. It is successfully decoupling, and no amount of additional strategy will fix that. 4.4 Integration Taken together, the three lenses generate the central proposition of this article. Effective adoption of agentic AI requires dynamic capabilities that most administrations have not built; those capabilities require an ambidextrous structure that most administrations have not designed; and the pressure to appear modern encourages substitutes for both, in the form of symbolic adoption that satisfies institutional expectations while changing nothing. Administrative agility is the deliberate refusal of that substitution. 5. Method 5.1 Research design This study is an #integrative_review, a method appropriate where the aim is to synthesise a fragmented and fast-moving body of work and to build a conceptual framework rather than to test a hypothesis. It does not claim to be a systematic review in the strict sense, because the phenomenon under study is too recent for the peer-reviewed literature to have stabilised, and a strict protocol would exclude the most current material while including a large volume of work on earlier technologies. 5.2 Sources and selection Literature was drawn from three fields. From public administration and digital government: work on AI adoption in government, algorithmic decision-making, administrative discretion, and public sector innovation. From strategic management: work on dynamic capabilities, ambidexterity, and strategy under uncertainty. From information systems: work on the management of AI, value creation from machine learning, and human-AI interaction. The #inclusion_criteria were as follows. Sources were peer-reviewed journal articles, scholarly books, or authoritative institutional reports. Priority was given to work published between 2021 and 2025. Older work was included only where it establishes a foundational concept still in active use. Work concerning purely technical model architecture, with no organisational or governance implication, was excluded. Vendor material and promotional writing were excluded as evidence, although they are discussed in the analysis as objects of study, since they shape what administrators believe. 5.3 Analysis Selected sources were read and coded thematically. The #thematic_analysis proceeded in three passes. The first pass identified the claims each source made about the relationship between organisational stability and technological change. The second grouped these claims into candidate themes. The third tested the candidate themes against the theoretical framework, retaining those that were supported by more than one independent source and that remained meaningful when applied specifically to agentic rather than predictive or generative systems. Six themes survived this process. They are presented in Section 7. 5.4 Limitations of the method Three limitations should be stated plainly. First, the peer-reviewed evidence base concerning agentic systems specifically, as opposed to AI in general, is thin, because deployments are recent. Much of the analysis therefore extends findings from adjacent technologies, and this extension is an argument, not a proof. Second, the #conceptual_synthesis reflects the author's judgement in weighing sources, and a different reviewer might weigh them differently. Third, the literature is heavily skewed toward European and North American administrations, which limits the transferability of conclusions to administrations operating under different legal traditions, resource levels, and state capacities. These limitations are revisited in Section 12. 6. The Core Tension: Speed Against Stability 6.1 Stating the paradox The #speed_stability_paradox can be stated simply. The value of a strategy comes from commitment: from the fact that resources are aligned behind it, that people can plan on the basis of it, and that it does not change with every passing enthusiasm. The value of agility comes from the ability to abandon a course of action quickly when it stops working. Each of these values undermines the other. Commitment without the ability to abandon produces #strategic_drift, in which an organisation continues to execute a plan whose premises have failed. The ability to abandon without commitment produces churn, in which nothing is executed long enough to produce a result. Agentic AI sharpens the paradox in three ways. 6.2 The compression of the decision cycle #technological_uncertainty in this domain is unusual in that it resolves quickly but never fully. Every few months, capabilities that were unavailable become available and capabilities that were expensive become cheap. This means that the correct answer to "should we automate this process?" may change within a single budget cycle. Administrations whose decision cycles are annual are structurally unable to keep pace, not because their staff are slow but because their governance rhythm is slower than the phenomenon it governs. The naive response is to speed up the governance rhythm, holding review boards monthly instead of annually. This helps slightly and creates a new problem: decision fatigue, and a governance body that reviews so many small items that it cannot see the pattern. The better response, developed later, is to change what is decided at each level. Strategic bodies should decide slowly about goals, risk tolerances, and red lines. Operational bodies should decide quickly about tools within those boundaries. The mistake is asking a strategic body to approve tools. 6.3 The irreversibility trap Some agentic adoption decisions look small and are large. Embedding an agent into a citizen-facing service changes the citizen's expectation of response time permanently. Once claims are answered in minutes, an administration cannot return to answering them in weeks without being seen to have failed. #sunk_costs also accumulate quietly: staff are retrained, processes are redesigned around the agent's output, and the old manual capability atrophies. Within two years, the "pilot" cannot be switched off, because nothing else can do the work. The related risk is #vendor_lock_in. Agentic frameworks are not commodities. Each has its own way of describing tools, storing state, and handling failure. An administration that builds fifty agents inside one proprietary framework has not bought a tool; it has changed its architecture, and the cost of changing supplier now includes rewriting all fifty. This is one of the clearest points at which a decision made for reasons of speed destroys future agility. Contractual and architectural insistence on portability at the start is cheap. Extracting oneself later is not. 6.4 Political windows and the temptation of announcement Administrations do not operate on technical logic alone. #policy_windows open when political attention, available funding, and a defined problem coincide, and they close again. An administrator who sees an opportunity to secure funding for AI capability during a window of enthusiasm is behaving rationally when they seize it, even if the technology is not yet mature for their use case. The danger is that funding secured through a window carries expectations attached to that window: visible results, quickly, and preferably before an election or a budget review. This pushes organisations toward the projects that demonstrate well rather than the projects that matter. Public-facing chatbots demonstrate well. Data infrastructure does not. The result is a systematic bias in the direction of investment, away from the foundations that determine long-term capacity and toward the surfaces that produce announcements. Recognising this bias is the first step to correcting it, and correction usually means using visible projects deliberately as a vehicle to fund invisible ones. 6.5 Reframing the paradox The way out of the paradox is to notice that it rests on an equivocation. "Strategy" is being used to mean two different things: the ends the organisation is committed to, and the means by which it pursues them. Confusing them is the actual error. The ends of an administration change slowly and should. A tax authority exists to collect revenue fairly and efficiently. A health service exists to improve health. A municipality exists to provide liveable conditions for residents. These purposes are not disturbed by a new model release. The means should change quickly, because they are instruments and nothing more. An organisation that has clearly separated the two can absorb enormous technical change without any strategic disturbance at all, because each new tool is evaluated by a stable question: does this help us collect revenue more fairly and efficiently? Agility, then, is not the enemy of strategy. Agility is what strategy looks like when the means are volatile and the ends are not. The organisations that experience new technology as a strategic crisis are usually organisations that had confused a particular method with their purpose. 7. Six Pressure Points The thematic analysis identified six areas where the collision between speed and stability is most consequential. 7.1 The accountability gap Administrative law rests on the principle that a decision affecting a person can be attributed to a decision-maker who can be required to explain and justify it. An agentic system disturbs this in a specific way. Its actions arise from a sequence of internal steps, each of which depends on the last, and the sequence is not fully specified in advance by any human. The organisation can say who deployed the system, but it cannot always say who decided the particular thing that was decided. This is the #accountability_gap. It is not solved by saying that a human remains formally responsible, because formal responsibility without practical control is a fiction, and a fiction that will not survive contact with a court or an ombudsman. Nor is it solved by #explainability tools alone. Explanations generated after the fact may be plausible, useful, and still not be a true account of why the system acted as it did. Two practical responses are more promising. The first is to constrain the action space so that whatever the system reasons internally, the set of actions it can take is small, enumerated, and individually reviewable. The second is to maintain complete #audit_trails at the level of action rather than reasoning: a permanent, tamper-evident record of every external call the agent made, with inputs and outputs, so that a decision can be reconstructed even if the reasoning cannot be inspected. #algorithmic_transparency in an agentic context is more achievable at the level of what was done than at the level of why. 7.2 Data infrastructure and the myth of readiness Agentic systems are more sensitive to data quality than their predecessors, not less. A predictive model with poor data produces a bad score, which a human may catch. An agent with poor data takes a bad action, and then takes further actions on the assumption that the first one was correct. Errors do not stay still; they propagate. #data_governance is therefore not a precondition that can be postponed. Yet in most administrations the true state is well known and rarely stated: records are duplicated across systems, definitions differ between departments, key fields are free text, and #interoperability between #legacy_systems is achieved by manual re-entry. #data_quality problems of this kind cannot be solved by an AI project. They must be solved before one, or the project will simply reveal them at high speed and in public. There is an uncomfortable implication here for the sequencing of investment. The most valuable AI work an administration can do in its first year may involve no AI at all. It may consist of building a service catalogue, defining canonical data models, creating documented interfaces to core registers, and establishing a single record of what systems exist and who owns them. This work is difficult to fund, because it produces no announcement. It is also the difference between an organisation that can adopt in year three and one that cannot. 7.3 Discretion and the frontline Street-level administrators have always exercised #administrative_discretion. Rules are general and cases are particular, and the gap between them is filled by human judgement. This discretion is a source of both flexibility and injustice, and administrative systems have long tried to channel rather than eliminate it. Agentic systems reshape discretion in a way that is easily missed. They do not simply remove it. They relocate it. Discretion migrates from the case officer, who exercised it visibly and could be questioned about it, into the design of the system: into the prompts, the tool permissions, the escalation thresholds, and the default behaviours. These design choices are typically made by technical staff and vendors, who do not think of themselves as making policy and are not accountable as policymakers. This relocation is the single most under-examined governance issue in the current wave. It means that decisions with real distributive consequences, about who gets escalated to a human and who does not, about what counts as a complete application, about when a case is closed, are being made in configuration files. An administration serious about agility must bring these choices back into the open, treating agent configuration as an instrument of policy that requires the same review as a written guideline. 7.4 Workforce, skills, and change fatigue The received wisdom is that staff resist AI because they fear replacement. The evidence suggests something more nuanced. Resistance is strongest where staff perceive the technology as unreliable, as imposed without consultation, or as a way of shifting blame to them when it fails. Staff who are given tools that visibly reduce tedious work, with authority to override, and without a threat to their position, generally adopt them. Two workforce issues deserve particular attention. The first is #workforce_reskilling, which is usually framed too narrowly as technical training. The scarce skill in an agentic administration is not prompt writing. It is the ability to specify a process precisely enough to delegate it, to design an escalation rule, to evaluate whether an agent's output is correct, and to notice when it is subtly wrong. These are analytical and supervisory skills, closer to the work of an auditor or a manager than a programmer. The second is #change_fatigue. Many administrations have already been through several waves of digital transformation, each announced as fundamental, each partly abandoned. Staff who have watched three previous systems arrive and depart have learned, rationally, to wait. This accumulated scepticism is an asset that has been destroyed and must be rebuilt, and it is rebuilt only by finishing things. An organisation that completes two modest agentic deployments, evaluates them honestly, and keeps the ones that work, will earn more #capability_building credibility than one that announces ten. 7.5 Procurement, dependency, and the shape of the market Procurement is where strategy either becomes real or becomes decorative. #procurement_reform is discussed endlessly in the public sector and is genuinely difficult, because the rules exist for good reasons: to prevent corruption, ensure fair competition, and protect public money. The problem is that these rules assume a stable specification. The buyer describes what is needed, suppliers bid, the best bid wins, and a contract is signed for a defined deliverable. Agentic technology does not fit this shape. The specification cannot be fixed in advance, because what is possible changes during the procurement period itself. Long procurement cycles therefore guarantee that administrations buy the previous generation of technology. Several adaptations are available and are being used: framework agreements that pre-qualify a pool of suppliers and allow rapid call-off; outcome-based specifications that define the result rather than the method; staged contracts with defined exit points; and mandatory portability requirements that require the supplier to deliver configurations and data in open formats. None of these is a complete answer. Together they can shift the balance of power back toward the buyer, which is where the public interest lies. The alternative is a market in which a small number of suppliers hold the operating knowledge of the state. 7.6 Legitimacy and public trust Finally, and most importantly, there is #legitimacy. An administration that loses the confidence of the public it serves has lost the thing that made it effective, and no efficiency gain compensates for that. #public_trust in algorithmic administration is fragile and asymmetric. It is built slowly through consistent, comprehensible, correctable decisions, and it is destroyed quickly by a single well-publicised failure, particularly one that harms people who were already vulnerable and who could not get the error fixed. Several documented failures in automated administration have followed exactly this pattern: a system that operated at scale, an error that was systematic rather than random, an appeal process that was slow and did not believe complainants, and eventually a public scandal that damaged the institution far beyond the original programme. The lesson is not that automation is dangerous. It is that scale is dangerous without correction. #equity concerns arise from the same source: an agent applies a consistent rule to everyone, which is fair in one sense, and which systematically disadvantages people whose circumstances do not fit the rule, which is unfair in another. Human discretion at least allowed for the exception. A well-designed agentic system must therefore build the exception path deliberately, and must make it fast, because an appeal process that takes six months against a decision that took six seconds is not an appeal process. It is an insult with paperwork. Related to this are #privacy and #cybersecurity. Agentic systems, by design, hold credentials and act on live systems. An agent that can be manipulated through the content it processes becomes a route into the organisation's core registers. #risk_management for agentic deployment therefore has a security dimension that predictive analytics never had, and it must be addressed before deployment rather than after the first incident. 8. The Anchored Agility Framework The six pressure points are not solved individually. They interact, and an organisation that addresses one while ignoring the others will fail. The framework proposed here is intended as an integrated response. It has five pillars. 8.1 Pillar one: strategic anchoring #strategic_anchoring is the discipline of separating ends from means and writing them down separately. In practice, an administration should maintain a short document, ideally no more than a few pages, stating what it exists to achieve, whom it serves, what it will never do regardless of technical possibility, and how it will know if it is succeeding. This document should be difficult to change and should be approved at the highest level. It should not mention any specific technology, because technologies are means. Beneath it sits a separate and explicitly disposable layer: the current portfolio of tools, systems, and methods. This layer is expected to change frequently. Changing it is a management decision, not a strategic event, and does not require reopening the anchor document. The value of this separation is that it removes the need for a "strategy" every time the technology moves. When a new capability appears, the question is not "must we rewrite our strategy?" but "does this help us do what we already said we exist to do, within the limits we already set?" That question can be answered in a meeting rather than a year. The red lines matter especially. An administration that decides in advance, calmly, that certain decisions will never be taken without a human, that certain data will never be sent outside its control, and that certain categories of case will always be escalated, has given itself a stable foundation from which to move fast on everything else. Red lines are not obstacles to agility. They are what makes agility safe. 8.2 Pillar two: modular adoption #modular_adoption means designing every agentic deployment so that its components can be replaced independently. Concretely: the reasoning model should sit behind an internal interface, so that changing model or supplier requires changing one component rather than fifty. Tools available to agents should be defined once, centrally, with permissions attached, and reused. Orchestration logic should live in the organisation's own code or configuration, not inside a vendor's proprietary environment. Data should remain in the organisation's stores, with the agent given access rather than possession. #reversibility is the design goal. Before any deployment, the organisation should be able to state precisely how it would be switched off, what would happen to the work in progress, and how long the reversal would take. If nobody can answer these questions, the deployment is not ready, whatever its technical merits. This pillar also implies a #portfolio_management approach rather than a project approach. The organisation should hold a visible portfolio of agentic deployments, each with a stated purpose, a cost, an evaluation, and a status. Deployments that do not earn their place should be closed. This sounds obvious and is rare: most administrations can list what they have started and very few can list what they have stopped. 8.3 Pillar three: layered guardrails #guardrails should be applied at four layers, matching the four layers of the agentic system described in Section 3. At the reasoning layer, controls concern which models may be used, where they run, and what data may be sent to them. At the tool layer, controls concern what actions an agent may take. This is the highest-leverage layer. Every tool should be classified by consequence: read-only, reversible write, irreversible write, external communication, financial effect. Agents should receive the minimum permissions required, and irreversible or financial actions should require explicit human authorisation by default. At the orchestration layer, controls concern limits on the loop: maximum steps, maximum spend, timeouts, and mandatory stopping conditions. An agent that cannot complete its goal must stop and escalate, not continue trying indefinitely. At the supervision layer, controls concern the human relationship. The organisation should define, for each deployment, whether the human approves each action, reviews a sample, or is notified only on exception, and this choice should be justified by the consequence of error rather than by convenience. Regulatory #regulatory_compliance obligations, including those emerging from risk-based AI regulation, map onto these layers reasonably well, and organisations that build the layered structure will find compliance easier than those that treat it as a separate documentation exercise. 8.4 Pillar four: institutionalised learning loops #learning_loops are the mechanism by which the organisation converts experience into capability rather than into anecdote. Three loops are needed, operating at different speeds. The fast loop is operational: #continuous_monitoring of agent behaviour, with automatic alerts when outputs drift from expected distributions, error rates rise, or escalation volumes change. This loop runs continuously and is largely automated. The middle loop is evaluative. Every deployment should have defined #evaluation_metrics agreed before launch, covering not only efficiency but quality, equity, and citizen experience. Measuring only speed guarantees that quality will quietly degrade, because speed is what is being watched. Evaluation should be conducted by a body that did not build the system. The slow loop is strategic. Once or twice a year, the organisation should examine its whole portfolio and ask what it has learned: which categories of task turned out to be suitable for delegation, which did not, where the assumptions in the anchor document are being strained, and what capability gaps have appeared. This is the loop that converts a collection of projects into an organisational capability. #incident_reporting deserves separate mention. Administrations should maintain a register of agentic incidents, including near-misses, and should treat it as a safety instrument rather than a disciplinary one. The aviation and healthcare sectors learned long ago that incidents are only reported where reporting is safe, and that unreported incidents recur. 8.5 Pillar five: legitimacy by design #legitimacy_by_design means that the acceptability of a system to the people it affects is treated as a design requirement from the beginning, not as a communications problem afterwards. This has several components. Citizens should be told when they are interacting with an autonomous system, in plain language, without having to search for the disclosure. There should be a clear, fast, and genuinely available route to a human, and using it should not disadvantage the person who uses it. Decisions should be explainable in terms the affected person can understand, which means the operative reasons, not a technical description of the model. Errors should be correctable quickly, and the burden of proof should not fall entirely on the person harmed. #co_production offers a stronger version of the same principle: involving affected groups, including staff and citizens' representatives, in the design of the system rather than in its evaluation after launch. This is slower at the start and much faster overall, because it surfaces objections while they can still be addressed cheaply. Attention to the #digital_divide belongs here too. Agentic services are excellent for people who are comfortable with digital interaction and can articulate their situation clearly. They are worse for people who are not, and those people are disproportionately the ones who most need the service. An administration whose automated channel works beautifully while its human channel decays has not improved its service. It has improved its service for the people who needed it least. 9. A Maturity Model for Anchored Agility The framework describes what a mature organisation looks like. Most organisations are not there, and the honest question is what to do next. The following #maturity_model sets out five stages. The stages are cumulative: skipping one does not accelerate progress, it merely postpones the failure. Stage 0: Symbolic The organisation has an AI strategy document, possibly a governance board, and a small number of pilots run by enthusiasts. Nothing has changed in any core process. Vendor demonstrations are frequent. Nobody can say what the pilots cost or whether they worked. This stage feels like progress and is not. The characteristic diagnostic is that the organisation cannot name a single process that now runs differently. The move out of Stage 0 is to stop announcing and start counting. Produce an honest inventory: what has been deployed, what it cost, what it does, who owns it, whether it is used. Stage 1: Foundational The organisation invests in what agentic systems will require: a documented catalogue of its own processes, defined data ownership, cleaned core registers, interfaces to legacy systems, and a basic security posture for machine access. It also establishes the anchor document with red lines, and a decision framework specifying who may approve what. This stage is unglamorous and politically difficult, because it produces little visible output. It is also where the organisations that eventually succeed spend their first year. The temptation to skip it is very strong and should be resisted. Stage 2: Contained deployment The organisation deploys agentic systems in areas where the consequence of error is low and the ability to detect error is high. Good candidates share three properties: the task is internal rather than citizen-facing, the output is checked by a human before it has effect, and the correct answer is knowable. Drafting, summarising, first-pass triage, reconciliation of records, and search across internal documents are typical. The purpose of this stage is not efficiency. It is learning. The organisation is discovering how these systems fail, what its staff need, whether its monitoring works, and whether its guardrails hold. Efficiency gains at this stage are a bonus, not the objective, and organisations that treat them as the objective will push into higher-risk deployment before they are ready. Stage 3: Governed autonomy The organisation extends agentic systems to consequential processes, with the full guardrail structure in place: action-level permissions, audit trails, escalation rules, evaluation by an independent body, an incident register, and a working appeal route for citizens. Human supervision is calibrated to consequence rather than applied uniformly. The distinguishing feature of Stage 3 is that the organisation can answer, for any given decision, who was accountable, what the system did, why, and how a person affected by it can challenge it. Organisations that cannot answer these questions have not reached Stage 3, regardless of how many agents they run. Stage 4: Anchored agility The organisation has an internal capability that can evaluate a new technical development, decide within weeks whether it is relevant, deploy it within a governed structure, evaluate it honestly, and retire it when it stops earning its place. The anchor document remains stable while the portfolio churns beneath it. Technical change is absorbed as routine management rather than experienced as crisis. Very few administrations are at this stage. It is worth stating clearly that Stage 4 is not a technology achievement. It is an organisational one, and it is reached by building #governance_board structures, decision rights, and evaluation habits, most of which would be valuable even if agentic AI had never been invented. 10. Patterns from Practice The literature on public sector AI documents several recurring patterns that illustrate the argument above. These are presented as patterns rather than as case evidence for specific claims, because the deployments described are recent and their evaluation is incomplete. 10.1 Conversational interfaces Citizen-facing #chatbots are the most common public sector deployment and among the least strategically significant. They are visible, they are relatively safe, and they demonstrate modernity. They also frequently disappoint, because the questions citizens actually ask are specific, situational, and depend on the state of their individual case, which the chatbot cannot see. The pattern worth noting is this: chatbot projects that succeed usually succeed because they force the organisation to write down its own rules clearly for the first time. The value came from the documentation, not the bot. This is a useful illustration of the general point that the preparation is often worth more than the artefact. 10.2 Triage and case routing #case_management triage is a more promising application. Incoming requests are classified, incomplete ones are identified, and cases are routed to the appropriate team. The consequence of error is moderate, the error is detectable, and the volume is high enough that gains are real. The governance issue is the one identified in Section 7.3: the classification rules embed policy. A triage system that consistently routes certain kinds of case to a slower queue has made a distributive decision, and that decision should be visible and reviewable, not buried in a configuration. 10.3 Predictive systems in enforcement #predictive_analytics used in enforcement, policing, and fraud detection has produced the most serious documented harms and the most valuable lessons. The recurring pattern involves a model trained on historical enforcement data, which reflects historical enforcement patterns rather than underlying behaviour; a resulting concentration of attention on populations already over-represented in that data; and a feedback loop in which the concentration generates further data confirming the original pattern. Adding agentic capability to such a system does not fix this. It accelerates it. The lesson for agility is that speed is only a virtue when the direction is right, and that the sensing capability described in Section 4.1 must include the capacity to notice when a system is efficiently doing the wrong thing. 10.4 Internal administrative work The least discussed and possibly most valuable application is internal: drafting, summarising, checking documents for consistency, preparing briefing material, and reconciling records across systems. #service_delivery improvements here are indirect but real, because the time recovered from administrative overhead returns to the substantive work. This category rarely appears in strategy documents, because it does not make an announcement. It is, for many administrations, where the honest return currently lies. 10.5 What the patterns suggest Across these patterns, three regularities appear. First, the deployments that create the most value are usually not the ones that create the most attention. Second, the binding constraint is almost never the model; it is the data, the process documentation, and the willingness of the organisation to change how work is done. Third, the organisations that report the best results are those that started smaller and evaluated more honestly than their peers, which is exactly what the maturity model in Section 9 would predict. 11. Discussion 11.1 Theoretical implications The analysis supports an extension of the dynamic capabilities perspective for public settings. In the private sector, sensing, seizing, and reconfiguring are directed at competitive advantage. In administration, they must be directed at #public_value, which includes but is not limited to efficiency, and which explicitly includes procedural legitimacy. A public dynamic capability is therefore not simply the ability to change quickly; it is the ability to change quickly in ways that remain justifiable to those affected. This adds a constraint that private-sector theory does not model, and it means that a public organisation that has become extremely fast at deploying technology may have become worse, not better, at its actual function. The analysis also suggests a refinement of ambidexterity theory. The conventional structural solution, an innovation unit separated from operations, is poorly suited to agentic AI, because the value of agentic systems lies precisely in their integration with operational processes and data. An isolated innovation unit will build impressive demonstrations that cannot be deployed. The alternative developed here, a thin central capability providing shared infrastructure and guardrails with distributed experimentation inside operational units, is closer to what the literature calls a platform model, and it deserves direct empirical study. Finally, institutional theory's concept of decoupling emerges as the central diagnostic risk of this technological wave. The pressure to appear modern is unusually strong, the technology is unusually easy to demonstrate and unusually hard to evaluate, and the gap between symbolic and substantive adoption is therefore unusually wide. Researchers studying #public_sector_innovation in this area should be sceptical of self-reported adoption and should look for evidence of changed process, not changed policy. 11.2 Practical implications For administrators, several implications follow directly. Separate the ends from the means, in writing, and defend the separation. Most of the anxiety attached to technological change in administration comes from the failure to make this separation, which turns every technical development into an existential question. Set red lines early, when the pressure is low. It is far easier to decide calmly, in advance, that certain decisions will never be automated than to hold that line under pressure from a minister and a supplier during a budget crisis. Invest in the invisible layer. The data catalogue, the process documentation, the register of systems, the interfaces: these determine what will be possible in three years, and nobody will thank anyone for building them. Buy for exit. Every contract should specify how the organisation leaves. If the answer is unclear, the price is higher than it appears. Evaluate honestly and publish the failures. An administration that closes a project because it did not work has learned something. An administration that quietly continues a project that did not work has learned nothing and is now paying for it. Protect the human channel. The measure of an automated public service is not how well it serves the people who find it easy. It is how well it serves the people who do not. 11.3 Implications for regulation The emergence of risk-based regulation, including the classification of certain public sector uses as high risk under the European #AI_act and comparable frameworks elsewhere, is broadly consistent with the argument advanced here. Regulation that focuses on the consequence of the action rather than the sophistication of the model matches the layered guardrail structure well. Two cautions are worth registering. First, compliance is not governance. An organisation can satisfy every documentation requirement and still deploy a system that harms people, if the documentation is produced by a compliance function that is separate from the people making design choices. Second, regulation aimed at models will age badly, because models change. Regulation aimed at actions, permissions, and remedies will age better, because those categories are stable. #responsible_AI in administration is ultimately about who can do what to whom, and what happens when it goes wrong. 12. Limitations and Future Research This article has clear limitations, which also mark the most useful directions for future work. The evidence base is immature. Agentic deployments in administration are recent, and the peer-reviewed literature necessarily lags. The framework presented here is therefore a structured argument built on adjacent evidence, and it requires #empirical_validation. The most valuable next step would be longitudinal case studies of organisations attempting agentic adoption, tracking not only outcomes but the sequence of decisions, including the ones that were reversed. The framework is untested against resource constraints. It assumes an administration with the capacity to build internal technical literacy, maintain a data catalogue, and conduct independent evaluation. Many administrations, particularly smaller municipalities and those in lower-income settings, have none of this. Whether the framework degrades gracefully under such constraints, or whether it simply becomes a description of what such organisations cannot do, is an open and important question. The geographic bias of the literature is significant. Most of the work synthesised here concerns European and North American administrations operating under particular legal traditions and with particular expectations of the state. #comparative_studies across different administrative traditions would test which elements of the argument are general and which are local. Several specific questions merit #future_research attention. How does administrative discretion actually relocate in practice when agentic systems are introduced, and who ends up exercising it? What organisational structures best support the hybrid ambidexterity model proposed here? What evaluation methods can detect subtle quality degradation in agentic outputs before it becomes a scandal? How should appeal processes be redesigned when the decision under appeal was taken in seconds? And what happens to institutional memory in an organisation whose routine work is increasingly performed by systems that do not remember? 13. Conclusion The question that opened this article was how an administration can hold a long-term course while absorbing a technology that changes faster than its planning cycle. The answer developed here is that the question contains a hidden confusion, and that clearing up the confusion dissolves much of the difficulty. An administration's purposes do not change because a new model was released. A tax authority still exists to collect revenue fairly. A health system still exists to improve health. These commitments are the anchor, and they should be written down, protected, and used as the standard against which every technical proposal is judged. What changes, and should change quickly, is the method. Treating a method as if it were a purpose is what causes organisations to experience technological change as a crisis of identity rather than a routine question of management. Administrative agility, then, is not speed for its own sake, and it is certainly not the abandonment of planning. It is a built capability with identifiable components: a stable anchor with explicit red lines; a modular architecture that keeps reversal possible; guardrails placed at the layer where consequences actually arise; learning loops that convert experience into knowledge; and a commitment to legitimacy that treats the acceptability of a system to the people it affects as a requirement rather than an afterthought. Agentic systems will make administrations that possess this capability considerably more effective. They will make administrations that lack it faster at doing the wrong thing, and they will make the resulting failures larger, more public, and harder to reverse. The difference between these two outcomes will not be determined by which models are purchased. It will be determined by work that looks, from the outside, entirely unrelated to artificial intelligence: cleaning data, documenting processes, clarifying who decides what, building an evaluation function that is willing to deliver bad news, and maintaining a route by which a citizen who has been wronged can be heard by a human being who has the power to put it right. That work is slow. It is also, precisely, what makes speed possible. Hashtags #Administrative_Agility #Agentic_AI_in_Government #Strategic_Alignment_in_Public_Sector #AI_Governance #Dynamic_Capabilities_Theory #Digital_Transformation_of_Government #Public_Sector_Innovation #Algorithmic_Accountability #Responsible_AI_Adoption #Organizational_Ambidexterity #AI_Strategy_for_Administrators #Human_Oversight_of_AI #Public_Value_and_Technology #Autonomous_Systems_in_Administration #STULIB_Research References Ahn, M. J., & Chen, Y.-C. (2022). Digital transformation toward AI-augmented public administration: The perception of government employees and the willingness to adopt AI in government. Government Information Quarterly, 39(2), 101664. Anthony, C., Bechky, B. A., & Fayard, A.-L. (2023). Collaborating with AI: Taking a system view to explore the future of work. Organization Science, 34(5), 1672–1694. Benbya, H., Pachidi, S., & Jarvenpaa, S. L. (2021). Special issue editorial: Artificial intelligence in organizations: Implications for information systems research. Journal of the Association for Information Systems, 22(2), 281–303. Berente, N., Gu, B., Recker, J., & Santhanam, R. (2021). Managing artificial intelligence. MIS Quarterly, 45(3), 1433–1450. Bharosa, N. (2022). The rise of GovTech: Trojan horse or blessing in disguise? A research agenda. Government Information Quarterly, 39(3), 101692. Bullock, J. B., Chen, Y.-C., Himmelreich, J., Hudson, V. M., Korinek, A., Young, M. M., & Zhang, B. (Eds.). (2024). The Oxford Handbook of AI Governance. Oxford University Press. Chan, A., Salganik, R., Markelius, A., Pang, C., Rajkumar, N., Krasheninnikov, D., Langosco, L., He, Z., Duan, Y., Carroll, M., Lin, M., Mayhew, A., Collins, K., Molamohammadi, M., Burden, J., Zhao, W., Rismani, S., Voudouris, K., Bhatt, U., Weller, A., Krueger, D., & Maharaj, T. (2023). Harms from increasingly agentic algorithmic systems. In Proceedings of the 2023 ACM Conference on Fairness, Accountability, and Transparency (pp. 651–666). Association for Computing Machinery. de Bruijn, H., Warnier, M., & Janssen, M. (2022). The perils and pitfalls of explainable AI: Strategies for explaining algorithmic decision-making. Government Information Quarterly, 39(2), 101666. Dwivedi, Y. K., Kshetri, N., Hughes, L., Slade, E. L., Jeyaraj, A., Kar, A. K., Baabdullah, A. M., Koohang, A., Raghavan, V., Ahuja, M., Albanna, H., Albashrawi, M. A., Al-Busaidi, A. S., Balakrishnan, J., Barlette, Y., Basu, S., Bose, I., Brooks, L., Buhalis, D., … Wright, R. (2023). Opinion paper: So what if ChatGPT wrote it? Multidisciplinary perspectives on opportunities, challenges and implications of generative conversational AI for research, practice and policy. International Journal of Information Management, 71, 102642. Floridi, L. (2023). The Ethics of Artificial Intelligence: Principles, Challenges, and Opportunities. Oxford University Press. Fountain, J. E. (2022). The moon, the ghetto and artificial intelligence: Reducing systemic poverty via fundamental change. Government Information Quarterly, 39(2), 101645. Helfat, C. E., Kaul, A., Ketchen, D. J., Barney, J. B., Chatain, O., & Singh, H. (2023). Renewing the resource-based view: New contexts, new concepts, and new methods. Strategic Management Journal, 44(6), 1357–1390. Kane, G. C., Nanda, R., Phillips, A. N., & Copulsky, J. R. (2021). The Transformation Myth: Leading Your Organization through Uncertain Times. MIT Press. Kelly, S., Kaye, S.-A., & Oviedo-Trespalacios, O. (2023). What factors contribute to the acceptance of artificial intelligence? A systematic review. Telematics and Informatics, 77, 101925. Madan, R., & Ashok, M. (2023). AI adoption and diffusion in public administration: A systematic literature review and future research agenda. Government Information Quarterly, 40(1), 101774. Maragno, G., Tangi, L., Gastaldi, L., & Benedetti, M. (2023). Exploring the factors, affordances and constraints outlining the implementation of artificial intelligence in public sector organizations. International Journal of Information Management, 73, 102686. Meijer, A., Lorenz, L., & Wessels, M. (2021). Algorithmization of bureaucratic organizations: Using a practice lens to study how context shapes predictive policing systems. Public Administration Review, 81(5), 837–846. Mergel, I., Dickinson, H., Stenvall, J., & Gasco, M. (2023). Implementing AI in the public sector. Public Management Review, 1–14. Mikalef, P., & Gupta, M. (2021). Artificial intelligence capability: Conceptualization, measurement calibration, and empirical study on its impact on organizational creativity and firm performance. Information & Management, 58(3), 103434. Mikalef, P., Lemmer, K., Schaefer, C., Ylinen, M., Fjørtoft, S. O., Torvatn, H. Y., Gupta, M., & Niehaves, B. (2022). Enabling AI capabilities in government agencies: A study of determinants for European municipalities. Government Information Quarterly, 39(4), 101596. National Institute of Standards and Technology. (2023). Artificial Intelligence Risk Management Framework (AI RMF 1.0). U.S. Department of Commerce. Neumann, O., Guirguis, K., & Steiner, R. (2024). Exploring artificial intelligence adoption in public organizations: A comparative case study. Public Management Review, 26(1), 114–141. Organisation for Economic Co-operation and Development. (2024). Governing with Artificial Intelligence: Are Governments Ready? OECD Artificial Intelligence Papers. OECD Publishing. O'Reilly, C. A., & Tushman, M. L. (2021). Lead and Disrupt: How to Solve the Innovator's Dilemma (2nd ed.). Stanford Business Books. Raisch, S., & Krakowski, S. (2021). Artificial intelligence and management: The automation–augmentation paradox. Academy of Management Review, 46(1), 192–210. Shollo, A., Hopf, K., Thiess, T., & Müller, O. (2022). Shifting ML value creation mechanisms: A process model of ML value creation. Journal of Strategic Information Systems, 31(3), 101734. Tangi, L., Janssen, M., Benedetti, M., & Noci, G. (2021). Digital government transformation: A structural equation modelling analysis of driving and impeding factors. International Journal of Information Management, 60, 102356. van Noordt, C., & Misuraca, G. (2022). Artificial intelligence for the public sector: Results of landscaping the use of AI in government across the European Union. Government Information Quarterly, 39(3), 101714. Vassilakopoulou, P., Haug, A., Salvesen, L. M., & Pappas, I. O. (2023). Developing human/AI interactions for chat-based customer services: Lessons learned from the Norwegian government. European Journal of Information Systems, 32(1), 10–22. Wirtz, B. W., Langer, P. F., & Fenner, C. (2021). Artificial intelligence in the public sector: A research agenda. International Journal of Public Administration, 44(13), 1103–1128. Wirtz, B. W., Weyerer, J. C., & Kehl, I. (2022). Governance of artificial intelligence: A risk and guideline-based integrative framework. Government Information Quarterly, 39(4), 101685. Zuiderwijk, A., Chen, Y.-C., & Salem, F. (2021). Implications of the use of artificial intelligence in public governance: A systematic literature review and a research agenda. Government Information Quarterly, 38(3), 101577.