Few questions in social argument are asked with more confidence and answered with less agreement than this one. One writer announces that the working class has vanished, dissolved into a broad middle mass of homeowners, graduates, and pension holders. Another replies that it has never been larger, that most of humanity now lives by selling its capacity to work, and that reports of the disappearance confuse a change of costume with a death. Both writers can point at real evidence. Both can quote employment statistics, income distributions, and scenes from ordinary life that support their side. Yet the argument never closes, because the two sides are not disagreeing about the same object. Each has quietly adopted a different test for what a working class is, and each test, applied honestly, delivers the verdict its holder announced in advance. The counting is downstream of the definition, and the definition is where the real dispute lives.

This article proceeds on a conviction that the dispute can be made intelligible, and largely resolved, by dragging its hidden machinery into the open. The machinery is a criterion: a rule that decides who counts as working class before any counting begins. State the criterion and the verdict follows almost mechanically. Leave it unstated and two intelligent people can argue forever while measuring different things. That conviction has a name in what follows. The criterion-first rule holds that any claim that the working class has disappeared or persisted is uninterpretable until the criterion is stated, and that once it is stated, most of this debate resolves into two groups measuring different things and disagreeing about which measurement matters. The rule does not decide the argument by itself. It shows where the argument actually is, which parts of it evidence can settle, and which part is a genuine contest about what class is for.

Does the working class still exist?

On Marx’s criterion, yes, and it has grown. Class is defined by the need to sell labour power, not by manual work or industry, so service, clerical, and technical employees qualify. What has shrunk is the industrial working class of the older economies, not wage dependence itself.

That short answer carries the whole structure of the longer one, so it is worth seeing what it commits its holder to. It says the question has a definitional hinge. It says the hinge is a relation to the means of production rather than a type of job, a level of pay, or a style of life. It says the famous shrinkage is real but located: a particular historical form of the class, concentrated in the factories and mines of the older industrial economies, contracted sharply across the closing decades of the twentieth century, while the underlying relation spread across services, offices, hospitals, warehouses, and continents. A reader who accepts the criterion gets this answer. A reader who rejects it gets another, and no pile of statistics, however carefully assembled, will reconcile them until the criterion itself is argued about directly. Most public argument skips that step, which is why most public argument on this question generates heat in exact proportion to the light it avoids.

The stakes are not merely academic. Class is a load-bearing word in political speech. Parties court it, journalists pronounce on its moods, and strategists announce its realignment or its irrelevance with each electoral cycle. If the word names a relation that still structures the lives of most people, then its alleged disappearance is a consequential error, a misreading of the society being governed. If the word names a vanished industrial world, then clinging to it is nostalgia dressed as analysis. Either way, the criterion does the deciding. A debate and controversy treatment owes the reader both sides at full strength, the dissolution case in the terms its best defenders chose, the persistence case with its concessions honestly registered, and a verdict that follows from the comparison rather than preceding it. That is the order this article keeps. First the criterion problem itself, then each criterion examined, then the great disappearance argument stated as strongly as it can be stated, then the evidence weighed, then the hardest objections faced without flinching.

Why the question is definitional before it is empirical

Begin with a simple experiment. Take one person, a nurse employed by a large hospital, and ask whether she belongs to the working class. Under a criterion of manual labour she does not, or only awkwardly, since her work mixes physical effort with clinical judgement, records, and machinery. Under a criterion of income she might fall either side of whatever band is chosen, depending on her grade and her shifts. Under a criterion of culture and self-identification the answer depends on her family, her neighbourhood, her accent, her own sense of where she stands. Under a criterion of authority she supervises nobody and takes instructions within a hierarchy, which places her low in one ordering. Under the relational criterion she qualifies without strain: she owns no hospital, no clinic, no means of earning a living except her trained capacity to work, and she must sell that capacity to an employer in order to live. Five criteria, five different maps, and the nurse migrates across the class boundary as the test changes. Nothing about the nurse changed. The instrument changed.

Multiply the experiment across a whole society and the scale of the definitional problem becomes clear. A population sorted by occupation gives one class structure. The same population sorted by income gives another. Sorted by property, by education, by self-description, it gives others again, and the overlaps between the maps are partial at best. Disputes about whether the working class has shrunk then turn out, on inspection, to be disputes between maps. One side exhibits the occupational map, on which factory employment collapsed, and declares the class dead. The other side exhibits the property map, on which the share of people who must work for a wage or salary rose across the same decades, and declares the class enlarged. Both exhibits are accurate. The disagreement is not about the exhibits. It is about which map deserves the name of class, and that is a theoretical disagreement wearing empirical clothes.

This is not a pedantic preliminary. It changes what would count as settling the question. If the dispute were purely empirical, better data would end it: a sufficiently detailed occupational census, a sufficiently long run of income surveys. Sociologists have produced both in abundance, and the dispute continues undisturbed, because each new dataset is simply read through the criterion its reader already holds. The factory census confirms disappearance to one reader and relocation to another. The wage statistics confirm persistence to one reader and prove nothing about class to another, who insists that class was always about culture and solidarity rather than pay packets. Data cannot arbitrate between criteria. Only argument about criteria can do that, and argument about criteria means asking what a class concept is supposed to explain. A criterion is vindicated, if it is vindicated at all, by the work it does: the patterns it predicts, the connections it reveals, the phenomena it renders intelligible that rival criteria leave as noise.

That standard will be applied throughout this article, and it should be applied to the relational criterion as severely as to any other. The relational criterion claims that a person’s position in the relations of production, above all whether they must sell their labour power to live, predicts a distinctive exposure: dependence on a labour market, subjection to managerial authority, vulnerability to dismissal and to the business cycle, and an interest, whether recognised or not, in the terms on which labour power is bought and used. If that package of predictions failed, if wage earners turned out to be as secure, as autonomous, and as insulated from economic weather as owners, the criterion would be an empty label and its defenders would be protecting a word rather than an analysis. The honest treatment, which the later sections attempt, asks where the package holds, where it strains, and what its defenders must concede. Rival criteria get the same treatment. The manual criterion predicts a world that deindustrialisation then removed, which is its weakness as analysis but also the source of its emotional truth. The income criterion tracks living standards well and class badly. The cultural criterion captures something real about identity and solidarity that the relational map misses entirely, and its defenders are entitled to say so. The criterion comparison table later in this article sets the rivals side by side so the reader can watch the same society change shape as the test changes.

One more distinction belongs in this opening groundwork, because confusion about it fuels half the quarrel. A class in the structural sense, a set of people occupying a common position, is not the same thing as a class in the political sense, a body of people aware of that position and organised around it. Marx’s own writing moves between the two, sometimes within a page, and his famous expectations concerned their eventual alignment: the structurally defined class would become a class for itself, conscious and active. The twentieth century tested that expectation severely. Structural position and political behaviour repeatedly failed to correspond in the ways the simpler versions of the theory suggested, and the failure is a genuine weakness that the persistence case must carry openly rather than explain away. But the two questions remain separable. Whether wage dependence still structures most lives is one question. Whether those who live by wages understand themselves as a class and act as one is another. The disappearance argument gains much of its force by sliding between them, exhibiting declining union density, weakening class voting, and fading class identity as though they were evidence about structural position. They are evidence about organisation and consciousness. A class can be structurally vast and politically quiet, and whether that combination refutes or merely complicates the theory is a matter the verdict section takes up directly.

The relational criterion: class as a position, not a costume

The criterion at the centre of the Marxist tradition defines class by a person’s relation to the means of production, the land, buildings, machinery, and organised capacities without which production cannot proceed. Those who own such means and live by employing others stand on one side. Those who own none, or none sufficient to live from, and must therefore sell their labour power to an owner stand on the other. Between them sits the older intermediate stratum of the self-employed, who own their own modest means of production and work them themselves. The full statement of this criterion, its textual basis in Marx and its later refinements, is set out in the series treatment of the marxist theory of class guide, and this article applies rather than rebuilds it. What matters here is what the criterion notices and what it deliberately ignores. It notices the compulsion to sell and the dependence that follows. It ignores, by design, the content of the job, the size of the pay packet, the cleanliness of the hands at the end of the shift, and the furniture in the worker’s home.

Marx described the wage worker as free in a double sense, and the phrase still does more work than any slogan in this debate. The worker is free, first, in the legal sense: not a slave, not a serf, bound to no master by status, entitled to sell labour power to whichever buyer will have it and to withdraw from any particular bargain. The worker is free, second, in the bleaker sense of being freed from the means of production: separated from land and tools sufficient to live independently, and therefore compelled, by the absence of alternatives rather than by any overseer, to find a buyer for the one commodity retained. Both freedoms are real, and the theory’s point is that they coexist. The labour market is genuinely a market, with genuine choice among employers, and it is also a structure of compulsion, since declining to sell to anyone at all is not among the choices. Every later refinement of the criterion is an attempt to say, for harder and harder cases, on which side of that compulsion a given person stands.

What does it mean to sell your labour power?

It means living by selling your capacity to work, for a wage or salary, because you own no means of production sufficient to support you independently. The sale is of labour power, the ability, rather than of a finished product, and it places the seller under an employer’s direction for the working period.

The test embedded in that answer is practical enough to apply to any biography. Could this person stop working for an employer and still live, at a tolerable standard, from property they control? If the answer is no, if the salary stopping means the mortgage, the rent, the food, and the heating stopping within months, then the person sells labour power whatever their job title says. If the answer is yes, if rents, dividends, or a saleable business would carry them indefinitely, then they stand elsewhere in the structure, whatever their habits or sympathies. Between the clear cases lie the instructive hard ones, the manager with a meaningful share portfolio, the senior professional with two rental flats, the tradesperson who owns a van and tools and works alone. The criterion does not pretend these cases are simple. It claims they are hard in an informative way, hard because two relations coexist in one household in proportions that can be estimated, and later sections return to them through the pension question and through Erik Olin Wright’s account of contradictory locations.

Notice what the criterion does with the cases that fill modern economies. The hospital nurse qualifies. So does the software developer employed by a firm whose code she does not own, the call centre agent reading a script, the delivery driver working to a platform’s schedule, the teacher in a state school, the junior lawyer salaried by a partnership she has not joined, the warehouse picker, the care worker visiting clients by the hour, and the airline pilot, whose pay is high and whose skill is scarce and who nevertheless owns no aircraft and must sell his labour power to fly. The criterion sweeps across the boundary between manual and non-manual work without pausing, because that boundary was never its concern. It also sweeps across the boundary between private and public employment, since a state employer buys labour power on the same structural terms, a point economists within the tradition have debated in connection with productive and unproductive labour without ever doubting that state employees who live by their salaries occupy the wage dependent position. And it handles the unemployed without embarrassment: a person between jobs, seeking a buyer for labour power and living meanwhile on savings, family, or public support, has not exited the relation but is experiencing its sharpest edge, the moment when the commodity cannot be sold at all.

Against this breadth, the rival criteria can now be stated fairly, because each captures something the relational map flattens. The manual criterion identifies the working class with those who work with their hands, above all in industry, mining, and transport, and it carries the authority of the class’s classic historical image: the factory hand, the miner, the docker. The income criterion identifies class with bands of earnings or wealth, locating the working class at the bottom of the distribution and the middle class across its broad centre. The cultural criterion identifies class with identity, speech, taste, education, and self-placement, the world a person feels they belong to and the world others assign them to. The authority criterion, developed most carefully inside the Marxist tradition itself, sorts people by their place in workplace command, separating those who direct the labour of others from those who only perform their own. The credential criterion, again Wright’s contribution in part, treats certified skill as an asset that complicates the simple ownership map, since scarce credentials give their holders bargaining power and rents of a kind that unskilled labour power cannot command. Each criterion is a genuine instrument. Each measures something real. The argument of this article is that they measure different things, that only one of them is the criterion Marx’s theory is built on, and that obituaries for the working class almost always arise from reading the results of one instrument as though they were the results of another.

The manual criterion deserves a moment of sympathy before criticism, because its grip on the imagination is earned. The industrial working class of the nineteenth and early twentieth centuries was not merely a statistical stratum. It was concentrated in great workplaces and dense districts, marked by shared dangers, shared timetables, and a shared adversary visible across the factory yard. Its organisations, unions, parties, clubs, and chapels, grew out of that concentration, and its culture was correspondingly thick and confident. When writers of the disappearance school describe what has been lost, it is this world they mean, and it has indeed largely gone from the older industrial economies. The error is one of identification. A historically specific form of the working class, industrial, manual, concentrated, and organised, is mistaken for the class itself, so that the form’s decline is reported as the relation’s extinction. On the manual criterion the mistake is invisible, because form and class are the same thing by definition. On the relational criterion the mistake is the whole story: the form changed beyond recognition while the relation, the need to sell labour power, extended itself over more people, in more sectors, in more countries, than ever before.

Six criteria, six different working classes

The comparison can be made concrete. The table below sets the main criteria side by side, showing who characteristically uses each, what the working class looks like under it, which way it has moved across the era of deindustrialisation in the older economies, and what the criterion is genuinely good for. Read down any single column and a coherent story appears. Read across the rows and the debate’s secret is exposed: the stories are about different objects.

Criterion	Who uses it	What the working class looks like under it	Grown or shrunk	What the criterion is good for
Ownership of means of production	Marx and the Marxist tradition	Everyone who must sell labour power to live, across manual, service, clerical, and technical work	Grown, in the older economies and far more worldwide	Explaining dependence on the labour market, exposure to dismissal and the business cycle, and the wage relation itself
Manual work	Popular usage, much journalism, the class’s classic image	Factory hands, miners, transport and construction workers	Shrunk sharply in the older economies	Describing the historical industrial form of the class and the world built around it
Income band	Policy discussion, marketing, electoral analysis	The lowest earning strata, shading into a broad middle	Shrunk as a share where real incomes rose; boundaries move with the bands chosen	Tracking living standards, poverty, and consumption
Cultural identification	Sociology of identity, political commentary	Those who describe themselves, and are described, as working class	Shrunk in self-identification across many surveys	Capturing belonging, solidarity, accent, taste, and the felt reality of class
Authority in the workplace	Wright and class analysis within the tradition	Those who perform labour without directing the labour of others	Roughly stable to grown, as hierarchies persist inside service and office work	Locating managers and supervisors, and the loyalties their command creates
Credential assets	Wright, labour market sociology	The unskilled and semi-skilled without scarce certified qualifications	Shrunk as a share as educational attainment rose	Explaining bargaining power, scarcity rents, and divisions among wage earners

Two features of the table repay attention. The first is that the growth and shrinkage question has no criterion-free answer. Ask whether the working class has shrunk and the manual row answers yes, the income row answers probably, the cultural row answers yes in the currency of identity, and the ownership row answers no, in the older economies and emphatically no across the world as a whole. The second feature is that no row is fraudulent. Each criterion is good for something, and the sociological literature that uses the cultural and income criteria is not confused about its own purposes. Confusion enters only when a result obtained on one row is announced in the vocabulary of another, when a shrinkage measured by hands or by self-description is reported as the disappearance of a class defined by property. That announcement, in its many forms, is the disappearance thesis, and it is time to state that thesis at the full strength its authors gave it.

The disappearance thesis at full strength

The case that the working class has ceased to exist, or ceased to matter, was not invented by careless journalists. It was built, over several decades, by serious thinkers responding to real transformations, and it deserves to be met in its strongest form rather than in the diluted version that is easy to defeat. Stated fully, it runs as follows. The social formation that Marx analysed, and that his theory of class describes, rested on a particular occupational structure: a large and growing industrial proletariat, concentrated, homogeneous in its conditions, and increasingly organised. That structure has been dismantled in the advanced economies. Manufacturing employment collapsed as a share of the whole. The great workplaces fragmented into networks of small firms, agencies, franchises, and contractors. Occupations diversified beyond any single description. Incomes and education rose across generations. Workers became homeowners, shareholders through their pension funds, and consumers courted by every party. Under these conditions, class in the old sense loses its grip three times over: the structural base shrinks, the cultural identity fades, and the political expression dissolves into cross-class appeals to the median voter. What remains is a gradient of lifestyles and incomes, not a structure of classes, and theory should follow reality into retirement.

The intellectual architecture of this case was assembled in stages. The post-industrial thesis, set out by Daniel Bell in the early 1970s, argued that the advanced economies were passing beyond industrial society altogether: theoretical knowledge rather than energy or machinery had become the axial resource, services were displacing goods production, and the professional and technical stratum was becoming central. Bell was no simple celebrant, and his argument was about the logic of the emerging order rather than the death of any class, but its implication for the classic proletariat was unmistakable: the social type on which Marx’s expectations rested belonged to a passing phase of production. Around 1980 André Gorz gave the implication its most famous literary form in a farewell to the working class, arguing that the industrial proletariat could no longer be the universal subject of history, that work itself was being reorganised in ways that dissolved the old identities, and that liberation would have to be imagined beyond the wage relation’s classic bearer. In the mid-1990s Jan Pakulski and Malcolm Waters pressed the argument to its theoretical conclusion in a thesis of the death of class itself: stratification, they argued, was increasingly organised by status, consumption, culture, and lifestyle rather than by production relations, so that class had been a historically bounded phenomenon, powerful in the industrial era and now yielding to other principles of social division.

Alongside these theoretical statements ran an empirical literature that gave the thesis its texture. Studies of affluent workers in the 1960s, centred on well-paid manual employees in new industries around Luton, tested the embourgeoisement hypothesis, the claim that prosperity was turning workers into members of the middle class in outlook and allegiance. The Luton researchers themselves reported a more complicated finding, instrumental collectivism rather than simple embourgeoisement, workers who treated unions as tools and affluence as a private project, but the broader lesson absorbed by the culture was that the old solidarities were thinning as living rooms filled with the apparatus of private comfort. Individualisation theory, associated with Ulrich Beck, added a further turn: the risks and biographies of late modern life are increasingly individualised, so that people experience structurally produced problems as personal successes and failures, and the collective categories through which those problems were once named lose their plausibility. Add the everyday observation available to anyone who travels through the older industrial regions, the closed pits, the converted mills, the docks become apartments, and the thesis acquires the force of common sense. Something unmistakably ended. The question is only what to call what ended, and on that naming everything depends.

The disappearance thesis also draws strength from politics, where its evidence is vivid. The great class parties of the industrial era either collapsed, moderated into catch-all electoral machines, or watched their traditional electorates fragment along lines of nation, education, age, and culture that the old map did not predict. Union density fell across most of the advanced economies across the closing decades of the twentieth century, and with it fell the institutional machinery that once made class a daily experience rather than an abstraction: the union branch, the labour club, the occupational community, the expectation that a person’s politics could be read from their job. A theory that predicted growing class organisation confronted a world of shrinking class organisation, and the disappearance writers were entitled to press the point hard. Their strongest formulation is not that workers became prosperous, though many did, nor that factories closed, though they did. It is that a category explains nothing if the people inside it share neither conditions, nor identity, nor organisation, nor political behaviour, and that the burden of proof lies on anyone who claims the category still binds. That formulation must be answered, not evaded, and the answer begins by separating the strands the thesis braids together: the occupational strand, the property strand, and the organisational strand. The occupational strand is sound. The organisational strand is largely sound. The property strand is where the thesis overreaches, and it is the strand the relational criterion exists to test.

What deindustrialisation actually changed

Strip the disappearance thesis down to its most-cited evidence and one fact does most of the work: in the older industrial economies, the share of employment in manufacturing and mining fell steeply across the final third of the twentieth century, in several countries from around a third of all jobs to a tenth or less. The fact is not in dispute, and the detailed evidence for it, together with the wider data on class structure, is assembled in the series companion on class structure evidence. What the fact means is everything. Read on the manual criterion, it records the working class draining away. Read on the relational criterion, it records the working class changing sectors, because the jobs that replaced the factory jobs, in retail, logistics, health, care, hospitality, cleaning, clerical work, and technical services, are overwhelmingly wage dependent jobs. The people who left the assembly line did not, in the main, become proprietors. They became employees of a different kind, and their children became employees in still larger numbers, as the routes into self-employment narrowed and higher education channelled graduates toward salaried professional work rather than toward independent practice.

Why did manufacturing employment mislead the argument for so long?

Because the factory was the class’s most visible historical form, its decline looked like the class itself declining. Manufacturing was concentrated, unionised, and culturally emblematic, while the service jobs replacing it were dispersed and less visibly collective, so an occupational contraction was misread as the disappearance of a relation.

Two further features of the occupational change deepen the misreading. The first is that manufacturing did not simply shrink; it also transformed internally, so that the boundary between making things and servicing them blurred beyond the reach of the old categories. A modern manufacturer employs designers, programmers, logistics coordinators, marketers, and maintenance contractors alongside a smaller number of production workers, and buys in cleaning, catering, transport, and computing from specialist firms whose staff then appear in the statistics as service workers although their labour serves the same production process as before. Outsourcing moves bodies between statistical sectors without moving them between class positions. The picker in a warehouse owned by a retailer’s logistics arm and the picker in a warehouse owned by a manufacturer are doing materially identical work under materially identical dependence; only the sector label differs. An argument built on sector shares will register a historic transformation where a relational analysis registers a relabelling, and much of the apparent disappearance dissolves under that single correction.

The second feature is subtler. Service work is not one thing, and its heterogeneity is constantly used against the persistence case, as though a category containing both a hospital consultant and a hospital cleaner must be meaningless. But heterogeneity of tasks never troubled the industrial class either, which contained the skilled fitter and the unskilled labourer, the miner at the face and the clerk in the colliery office, separated by pay, status, danger, and prospects as wide as anything separating service occupations now. What unified them was not similarity of work but similarity of position: dependence on the sale of labour power and subjection to an employer’s direction. Service employment reproduces that unity of position beneath an even greater diversity of tasks, which is why the relational criterion finds the class intact precisely where the occupational gaze finds only fragments. The point is developed further in the discussion of the gig economy below, since platform work presents the same unity and diversity problem in its sharpest contemporary form, and the series treatment of class analysis gig economy follows the argument into that contested territory.

The persistence case: wage dependence as the through-line

Set the relational criterion to work on the whole economy and a picture emerges that reverses the obituary. Across the advanced economies, the proportion of the working population who are employees, as against employers and the self-employed, rose across the very decades in which the working class was declared dead, reaching levels at which the overwhelming majority of those who work do so for a wage or salary paid by an organisation they do not own. The self-employed stratum that Marx treated as a substantial intermediate layer contracted in most older economies through the middle of the twentieth century, squeezed between large firms and salaried professionalism, before stabilising at modest levels that include many who are, in substance if not in legal form, dependent on a single client or platform. Agriculture, once the reservoir of independent producers, emptied into wage employment in town and city. The professions, once populated by independent practitioners, became salaried occupations practised inside hospitals, universities, firms, and departments. Each of these movements is a movement into the wage relation. The class defined by that relation did not shrink with the factories. It absorbed the society around the factories.

The character of the work performed inside the relation changed profoundly, and the persistence case loses credibility whenever it minimises that change. Service and office work brought different rhythms, different skills, different dangers, and different relations to the product than the factory imposed. A care worker’s labour is consumed in the act of its performance and leaves relationships rather than objects behind. A programmer’s output can be copied at no cost and her tools are rented from the same firms that employ her. A call centre agent’s effort is metered in seconds by systems a nineteenth-century overseer could not have imagined. None of this is the world of the mill chapters in Capital, and an analysis that pretended otherwise would be antiquarian. The claim is narrower and stronger: the mechanism that defines the class, the sale of labour power by those without sufficient property to live otherwise, operates through all of these settings unchanged in its essentials, and generates in each of them its characteristic pressures, the insecurity of the contract, the authority of the manager or the algorithm, the exposure to decisions taken elsewhere by owners the worker will never meet. Wherever those pressures are the ordinary texture of a working life, the relational map says the working class is present, whatever the work is called.

There is also a global dimension, and it transforms the argument’s scale. The disappearance writers wrote, in the main, about the older industrial economies, the North Atlantic world and its analogues, at exactly the moment when industrial production was expanding on a vast scale elsewhere. Across the closing decades of the twentieth century and into the twenty-first, East and South Asia industrialised at a pace and on a population base without precedent, drawing hundreds of millions from peasant agriculture and informal work into factories, construction, and urban services. The detailed magnitudes belong to the evidence literature rather than to this article, and they are best described directionally: the worldwide industrial workforce grew very substantially across the period in which obituaries for the working class were being written in London and New York, and the worldwide wage dependent population grew faster still, as urbanisation and the decline of subsistence production proceeded on every continent. The class did not vanish. It relocated, multiplied, and changed its centre of gravity, and the obituary was a local observation mistaken for a global one because the observers’ criterion tied the class to a landscape that capital itself was abandoning.

What happens to the count when manufacturing moves abroad?

It moves the count’s centre from the older industrial economies to East and South Asia, where industrial and urban wage employment expanded enormously as it contracted in the North Atlantic world. A criterion fixed on relations rather than places registers relocation and growth where a national occupational count registers only loss at home.

The relocation point deserves one refinement, because it is sometimes overstated into a simple story of jobs crossing borders. Industrial employment in the older economies also fell because productivity rose: fewer workers produced more goods, with the surplus absorbed by services at home rather than by factories abroad alone. Both mechanisms, automation and relocation, reduce the manual-industrial count in the old centres. Neither reduces the relational count. The worker displaced by a machine does not become a capitalist; she becomes an employee in a warehouse, a hospital, or an office, or an unemployed seller of labour power awaiting a buyer. The worker whose job moved abroad is joined, in the global total, by the worker abroad who now performs it under the same relation. Productivity growth and globalisation, the two great solvents of the industrial form, are both neutral or positive for the size of the class on the relational criterion. That is not a coincidence arranged for argumentative convenience. It follows from the criterion’s design, which tracks a relation that capitalist development itself generalises: the more completely production is organised through markets and firms, the more completely the population is sorted into those who own the firms and those who must work in them.

The property challenges: houses, pensions, and portfolios

The strongest evidence against the persistence case is not occupational at all. It is the spread of property among wage earners themselves. Across the postwar decades in the older economies, home ownership rose from a minority condition to the majority tenure; pension entitlements became near-universal among employees; and through pension funds and modest direct holdings, a large share of the workforce acquired some financial claim on the profits of capital. If the working class is defined by propertylessness, the objection runs, then the definition has been overtaken by the facts: the propertyless have become proprietors, and a criterion that still files a homeowner with a pension and an investment account alongside a nineteenth-century mill hand, who owned nothing but his hands, has stopped describing the world. This challenge deserves an argued answer rather than an asserted one, because it identifies real changes in workers’ material lives, and the reply succeeds only if it can explain why those changes, genuine as they are, do not move a person across the relational line.

Take the house first. A home is property, and for most households it is the largest asset they will ever hold. But the relational criterion never asked whether a person owns anything at all. It asked whether what they own amounts to means of production sufficient to live from without selling their labour power. A house occupied by its owner produces shelter, not income. It cannot be eaten, and its value can be realised only by ceasing to live in it or by borrowing against it, which converts the asset into a debt serviced from wages. The mortgaged homeowner is in fact the wage relation’s most tightly bound participant: the monthly payment disciplines the working life with a regularity no foreman could match, making the loss of employment an immediate threat to the home itself. Owner occupation changes the worker’s experience profoundly, softening some insecurities, adding others, and giving millions a stake in property values that complicates their politics. What it does not do is free its holder from the need to sell labour power, and on the operative test, the test of how a person lives, the homeowner who works for a wage remains inside the relation the criterion describes. The series discussion of the intermediate strata, petty bourgeoisie and middle class, examines the neighbouring case of the small proprietor who does own productive property, and the contrast is instructive: the shopkeeper’s asset works while he sleeps and can be sold as a going concern, which is exactly what the dwelling cannot do.

The pension case is harder, and the disappearance school is entitled to its full force. A pension fund is not a house. It is a vast pool of capital, invested in the shares and bonds of the very firms that employ the fund’s contributors, and in aggregate these funds became, across the late twentieth century, among the largest owners of capital in the advanced economies. Here, surely, the workers do own the means of production, collectively and increasingly. The reply must therefore be precise about what ownership means in the relational criterion, and the precision is available in the theory of property itself. Ownership, in the sense that matters for class, is a bundle whose decisive strand is control: the right to direct the asset, to appoint those who manage it, to decide what is produced and on what terms labour is hired. A pension contributor holds a claim on a future income stream, valuable, legally protected, and entirely passive. She does not choose the fund’s investments, cannot direct the firms it holds, cannot withdraw the capital to set up in business, and discovers her stake’s value only in statements she cannot influence. The fund’s managers, acting under mandates and market discipline, deploy the pool exactly as any other capital is deployed, pressing portfolio firms for returns with the same indifference to employment that the theory attributes to capital generally, sometimes against the interests of the very workers whose savings compose it. The result is one of the stranger structures of modern capitalism, workers’ deferred wages functioning as capital over workers, and it complicates any simple picture of two camps. But a claim is not control, and the operative test still answers cleanly: the pension holder who stops working before pension age discovers at once that the fund will not feed her in the present. She must sell her labour power to live. That is the relation, undisturbed by the size of the claim maturing in her name.

Do pensions and shares turn workers into capitalists?

No, because the decisive strand of ownership is control, and a pension or portfolio claim confers none. The holder cannot direct the firms held, appoint their managers, or live on the asset before it matures, so the need to sell labour power, the operative test of class position, remains exactly as it was.

Small direct shareholdings fall under the same analysis with less difficulty. A portfolio worth a few months of salary is a savings vehicle, not a means of production; its dividends season a wage rather than replace one, and its holder’s material life is governed by the labour market, not the stock market. The interesting cases lie further up the scale, where investment wealth grows large enough that its holder could plausibly live from it. Such people exist within salaried employment, senior executives above all, whose equity compensation can make them owners in substance while employees in form. The relational criterion handles them by proportion rather than by label: a person whose income flows mainly from property stands mainly in the owner’s position, whatever the contract says, and class analysis within the tradition has always been willing to classify by the dominant relation where two coexist. The existence of a gradient at the top of the salaried world does not dissolve the line for the great majority below it, for whom property income is trivial beside the wage and the wage’s cessation is an emergency. Lines in social analysis are rarely walls. A shoreline is real although the tide moves it, and the relational line is real although individual households cross it in both directions across a lifetime.

Education, credentials, and the second asset

The second great challenge is educational. Across the same decades, participation in higher education expanded from a small elite to a large minority and in some economies a majority of each cohort, and the graduate workforce became a central fact of the class debate. The educated employee, it is said, owns something after all: human capital, embodied, inalienable, and capable of commanding high returns, so that the propertyless worker of the theory has been replaced by a proprietor of skills whose market position resembles that of the petty bourgeois shopkeeper more than that of the mill hand. The vocabulary of human capital is mainstream economics rather than Marxism, but the challenge survives translation. If credentials function as an asset, does the relational map need a second dimension, and if it does, has the simple two-class picture given way to something the disappearance writers were right to sense?

Erik Olin Wright, the most systematic class analyst the tradition produced in its later decades, took this challenge seriously enough to rebuild the map around it. In his mature framework, exploitation and class rest on the unequal distribution of several kinds of productive asset: means of production in the classic sense, organisational assets embodied in authority over the production process, and skill or credential assets embodied in scarce certified capacities. Holders of scarce credentials can appropriate a rent, a premium above what their labour power would fetch unskilled, and that premium gives them interests partly aligned with owners, who also live by excluding others from an asset. The framework concedes to the educational challenge what is true in it. Credentials are a real asset in the limited sense that scarcity makes them valuable and exclusion makes them lucrative. A consultant surgeon or a senior software architect faces the labour market from a position of strength that a warehouse picker cannot imagine, and any class analysis that cannot register that difference is too coarse for honest use.

But an asset that must be carried into an employer’s workplace to yield anything differs in kind from an asset that yields while its owner sleeps. The credential cannot be sold separately from the person, cannot be bequeathed intact, cannot be left to accumulate while the holder takes a decade away, and perishes with the holder’s health and the fashion for the skill. Its rent is collected only through the sale of labour power, hour by hour, under an employer’s direction, and it vanishes the moment the sale stops. Wright’s own conclusion was not that credential holders form a separate class outside the wage relation. It was that they occupy locations within it that are contradictory, pulled toward labour by their dependence and toward capital by their rents, a refinement of the map rather than its abandonment. The educational expansion therefore changes the internal composition of the working class profoundly, stratifying it by credential, security, and bargaining power, and it changes the class’s culture and politics along with its composition. What it does not do is repeal the dependence that defines the class, since the graduate with a mortgage and a salary is bound to the labour market as completely as her grandfather was, and usually by larger debts. Expansion of education is, on the relational count, one more mechanism by which the wage relation generalised itself, absorbing occupations that once stood outside it as independent callings and reissuing them as salaried posts.

Contradictory locations: managers, supervisors, and the space between

Every large organisation confronts the class map with people who fit neither pole of the simple picture. The salaried manager owns no means of production and must sell her labour power, yet she hires, directs, disciplines, and sometimes dismisses other sellers of labour power on the owner’s behalf. The supervisor stands closer still to the line, the worker of one season holding the clipboard the next, exercising fragments of the employer’s authority while remaining exposed to the employer’s decisions. A map that files the chief executive and the cleaner in one class and stops there has plainly left something undescribed, and the disappearance writers were right to gesture at the swelling ranks of such intermediate figures as evidence that the old structure was dissolving into a continuum.

Wright’s answer, developed from the late 1970s onward, was the idea of contradictory class locations, and it remains the tradition’s most useful tool for this terrain. A location is contradictory when it shares the defining features of two classes at once, so that its occupants are, in Wright’s phrase, simultaneously in both. The manager is a wage labourer in relation to the owners above and a bearer of capital’s function in relation to the workers below: exploited in the formal sense that surplus is extracted from the organisation she serves, exploiter in the practical sense that she organises the extraction from others. The supervisor occupies a weaker version of the same duality, and the credentialed expert occupies another, combining dependence with scarcity rents. The full architecture of Wright’s framework and its empirical programme is treated in the series article on erik olin wright class analysis, and its importance here is what it does to the debate. It shows that the middle of the map can be elaborated indefinitely, with real explanatory gains, without surrendering the poles that give the map its orientation. A continuum with identifiable poles and theorised intermediate positions is not the absence of class structure. It is what a class structure looks like when a complex economy is mapped honestly rather than in the two tones of a pamphlet.

The intermediate positions also do explanatory work that the disappearance thesis needs and cannot supply from its own resources. If classes had simply dissolved into a gradient of incomes and lifestyles, the behaviour of managers in a crisis would be a matter of personal taste. On the contradictory locations account, it is predictable in outline: managers will defend organisational hierarchy and the prerogatives of direction even when their own salaries are squeezed, because their location ties their interests and their daily function to capital’s side of the relation, while their purely economic grievances, over pay, workload, and insecurity, pull episodically the other way. The long argument about the middle class, its composition and its loyalties, becomes tractable once this machinery is admitted, and the reader who wants the full taxonomy of the strata between the poles will find it in the treatment of the petty bourgeoisie and middle class already linked above. What matters for the existence question is the direction of the refinement. Every serious elaboration of the middle has been built by adding dimensions to the relational map, authority here, credentials there, which presupposes that the map’s base, the ownership relation, still organises the territory. One does not refine a map of a country that has ceased to exist.

The precariat: new class, or old condition renamed?

Into this long argument there arrived, in the early 2010s, a proposal that cut across it: Guy Standing’s thesis of the precariat. The proposal holds that globalisation and labour market flexibility have produced a new class in the making, distinct from the old working class, defined by insecure labour, unstable occupational identity, lack of the employment based protections the postwar settlement attached to jobs, and a distinctive consciousness of frustration and anxiety. The precariat, on this account, is not the proletariat under a new name. It is a successor formation, growing as the old class and its institutions decline, and politically volatile in ways the older labour movement never was. The thesis caught a real wind. Casual contracts, agency work, bogus self-employment, and platform labour did spread; occupational biographies did become more fragmented; the expectation of a job for life did retreat. Any answer that treats the precariat as a mere word trick fails the evidence as badly as the obituaries did.

The standard Marxist reply, which must be given at full strength alongside the thesis, is that the precariat describes a condition within the working class rather than a class distinct from it. Class, on the relational criterion, is defined by position in the relations of production, and the precarious worker’s position is the classic one in an exposed form: propertyless, dependent on selling labour power, subject to an employer’s or a platform’s direction. Insecurity is not a separate relation to the means of production. It is a variation in the terms on which the same relation is lived, and historically it is the older variation at that, since the secure, unionised, lifetime employment of the postwar decades was the exception in capitalism’s history rather than its norm. The dockers hired by the day, the seasonal hands, the outworkers, and the casual poor of the nineteenth-century city were precarious before the word existed, and nobody suggested they stood outside the working class for that reason. On this reply, the precariat thesis mistakes a change in the distribution of security inside the class for the birth of a new class, and its political predictions are correspondingly shaky, since a stratum defined by a condition it shares in varying degrees with millions of others may not cohere as a separate actor at all. The fuller exchange between the thesis and its critics is followed in the series article on precarity and the working class.

Is the precariat a separate class or a condition inside one?

A condition inside one, on the relational criterion. Precarious workers sell labour power and own no means of production, exactly as secure workers do; what differs is the stability of the contract and the protections around it. Insecurity grades the working class internally rather than marking a boundary around a new class.

Yet the reply should not be allowed to win too cheaply, because the precariat thesis registers two truths that the persistence case is tempted to skip. The first is that the internal stratification of the wage dependent population, by security, citizenship status, and access to public provision, can become so wide that shared position stops generating shared experience, and a class analysis that contents itself with the shared position will miss the lived reality of division, including the resentments that run horizontally among workers rather than vertically against owners. Standing’s portrait of a stratum angry at those above and those below, at the salaried and at the benefit recipient, captures something about the political chemistry of insecurity that the classic categories left unexamined. The second truth is that security itself was a historical achievement of class organisation, so that its loss is a fact about the class’s power and institutions, not merely about its composition. Where the reply is decisive is on the taxonomic question the existence debate actually turns on. A new condition, however politically potent, is not a new class unless it rests on a new relation to production, and no such relation has been produced. The platform driver who owns her car, sometimes advanced as the type specimen of a new formation, illustrates the point exactly: the vehicle is a means of production in the thinnest sense, while the means of access to customers, the platform, its dispatch system, its prices, and its power of exclusion, are owned elsewhere, and the driver’s dependence on that owned infrastructure is the relation that governs her working life. Form says proprietor. The operative test says seller of labour power with an unusually visible tool bill, and the criterion-first rule instructs us to trust the test over the form.

Boundary cases: the self-employed, the contractor, and the platform

Before the strongest objection is faced, the map’s edges deserve their own examination, because the disappearance argument often wins its audience at the edges, pointing at figures who seem to belong nowhere in the old scheme. The self-employed are the oldest such figures. The independent builder, the freelance designer, the corner shopkeeper, and the owner driver own modest means of production and work them personally, hiring little or no labour. Marx treated this stratum, the petty bourgeoisie, as a transitional class, squeezed between large capital and wage labour, and predicted its long decline as production concentrated. The prediction was directionally right for a century: small proprietors thinned as chains, platforms, and salaried professions absorbed their trades. Where self-employment persists or revives, it frequently does so in dependent forms that blur the old category from within. The contractor who works for one firm indefinitely, the courier classified as an independent business while taking prices and discipline from an application, the tradesperson whose order book is a single developer’s site: these figures hold the legal form of the petty bourgeoisie while living much of the substance of the wage relation, and classifying them requires the operative test applied earlier. Ask what governs the working life, whose decisions set its price and pace, and who bears the risk of idle time, and the answer usually assigns the person, whatever the tax form says, to dependence on a buyer of labour power with the costs of employment shifted onto the seller.

The blurring is real, and honesty requires saying what it costs the theory. A classification that must look through legal forms to substantive relations is harder to operationalise than one that reads the forms, and two analysts applying the substantive test to the same courier fleet can reach different counts. The traditions of research that followed Wright spent considerable effort on exactly this operational problem, coding occupations by ownership, authority, and credentials in large surveys, and the resulting estimates of class structure differ in their margins while agreeing in their broad shape: a small owning class, a large wage dependent majority, and intermediate strata whose size depends on the coding rules. That pattern, robustness of shape under variation of detail, is what a living research programme looks like, and it stands in instructive contrast to the cultural criterion, under which the working class’s measured size swings wildly with question wording, and to the manual criterion, under which the class’s size is hostage to the statistical office’s sector definitions. Every criterion pays for its insights somewhere. The relational criterion pays at its edges, in classification labour, and the payment is worth itemising because the strongest objection, to which this article now turns, is precisely the charge that the edges are not a cost but a refuge.

The strongest objection: a category that includes everyone explains nothing

The disappearance case has one formulation that rises above the rest, and the verdict of this debate depends on meeting it. Put at its best, it runs as follows. A category that places a consultant surgeon and a warehouse picker in the same class, while placing a self-employed builder who earns less than either outside it, is not tracking anything politically or sociologically real. The surgeon and the picker share no income, no status, no security, no daily experience, no culture, and in all likelihood no politics. The builder shares the picker’s income, neighbourhood, anxieties, and accent, and differs only in the legal form of his engagement. A classification that groups the strangers and separates the neighbours has inverted the purpose of social description. Worse, the classification is protected against refutation by its own method: whenever the evidence turns, the criterion is reasserted in a form that absorbs the blow. Factories close, and the class is found in offices. Workers buy houses and pensions, and ownership is redefined as control. Workers cease to identify or organise as a class, and consciousness is declared a separate question. A theory that survives every test by adjusting its definitions is performing the auxiliary hypothesis manoeuvre familiar from degenerating research programmes: the core is never at risk because the belt of reinterpretations takes every impact. On this charge, the relational criterion is not wrong. It is unfalsifiable, which for a claim about the world is worse.

The charge deserves a layered answer, and the first layer is a concession. The gap between structural position and political behaviour is real, it is large, and the theory’s early expectations about their alignment have not been borne out in the simple form in which they were held. Wage earners in the advanced economies have repeatedly voted against parties claiming to represent them as a class, have organised in falling proportions, and have described themselves in class terms less readily with each generation of survey research. Any persistence case that treats these facts as public relations problems, rather than as findings the theory must digest, deserves the auxiliary hypothesis charge. The honest position is that the relational map claims to describe structure and interest, not allegiance. It says the surgeon and the picker stand in the same relation to the means of production, and it predicts from that shared relation a specific and limited set of common exposures. It does not predict that they will feel themselves a class, vote as a class, or even recognise the description, and where Marx’s writing suggested the stronger prediction, the stronger prediction has fared badly and should be surrendered as a loss, recorded against the theory rather than reinterpreted away.

The second layer identifies what the relational category does predict that its rivals do not, because a criterion earns its keep in predictions or it earns nothing. The relational map predicts that the surgeon and the picker, for all their differences, are exposed to a common structure of risk that the builder, their apparent neighbour in income and culture, is not exposed to in the same form. Both can be dismissed. Both have their effort directed and monitored by an organisation they do not control, and both can have the terms of their work, its pace, its staffing, its very existence, altered by decisions in which they have no vote, taken by owners or by managers answerable to owners. The builder faces market risk, which is real and can ruin him, but he faces it as a seller of products, setting his prices and choosing his jobs, exposed to the anonymity of demand rather than to the authority of an employer. These are different vulnerabilities with different remedies: the employees’ vulnerability is addressed, where it is addressed, by employment law, collective bargaining, and organisational voice; the builder’s by competition policy, credit access, and prompt payment rules. A classification that sorts people by the kind of power they are subject to, and therefore by the kind of remedy their situation calls for, is tracking something a society ignores at its cost. Income bands cannot do this work, since the surgeon out-earns most builders by a factor that hides their shared exposure entirely, and the cultural criterion cannot do it, since identity follows accent and upbringing rather than the legal structure of a person’s dependence.

The third layer returns the falsifiability charge with interest, carefully. The relational criterion is falsifiable in the way structural claims generally are: it would fail if the exposures it predicts stopped tracking the ownership line, if employees as a body became as secure against dismissal, as autonomous in their work, and as insulated from their employers’ fortunes as the owners of the firms themselves. Nothing in the record of deindustrialisation shows that. What the record shows is the exposures migrating across sectors with the relation itself, turning up in the call centre’s monitoring software, the warehouse’s productivity regime, the hospital’s staffing ratios, and the platform’s algorithmic discipline, each a local dialect of the single language of directed labour. The adjustments the objection mocks are, on inspection, the criterion being held constant while the economy changed around it, which is the opposite of redefinition. The definition of the working class in terms of property and the sale of labour power is recognisably the one Marx used when the class was a minority of factory hands in a sea of peasants and artisans. It classified most of humanity as outside the class then. It classifies most of humanity as inside the class now, because the world moved toward the relation, not because the relation was restated to follow the world. A criterion that gave the unpopular answer in its author’s own century, when the proletariat was small and the peasantry vast, can hardly be accused of being tailored to flatter its later results.

One further prediction deserves naming, because it is the one the disappearance writers themselves rely on without noticing. The post-industrial and death of class theses were written as contributions to a recognisable genre, the announcement that a long-standing structure of power has dissolved into a flatter order of consumers, citizens, and knowledge workers. Each announcement arrived, by a coincidence its authors did not intend, in a decade when the power in question was being vigorously exercised: when unions were being defeated, employment protections thinned, and the share of income flowing to wages pressed downward against the share flowing to property. A sociology that had truly lost the category of class would struggle to describe those events except as weather. The relational map describes them in a sentence: the terms on which labour power is bought were revised in favour of its buyers, by organisation and law, against a class whose capacity to resist had been weakened by the very transformations the obituarists catalogued. That a weakened class is not an abolished class is the distinction on which this whole article turns, and it is a distinction the obituaries consistently refuse, because refusing it is what makes an obituary possible.

Did Marx expect this? The majority prediction and its fate

One strand of the debate concerns not the rivals’ criteria but Marx’s own expectations, and it is worth separating because both sides raid it for authority. Marx expected the working class to grow toward an overwhelming majority of society as capitalist development proceeded, with the intermediate strata ground down between the poles and production concentrating in fewer hands. The disappearance writers cite the middle part of that expectation, the polarisation, as having failed: the middle strata did not vanish, they metastasised, and the professions, managers, technicians, and administrators who fill them are the standing refutation of the two-class simplification. The persistence writers cite the first part, the growth toward a majority, as having succeeded beyond anything Marx saw: the wage dependent share of the world’s population is vastly higher than in his century, when most humans were peasants, and in that precise sense the proletarianisation of the world proceeded as the theory said it would. Both citations are accurate, and their coexistence is the lesson. The expectation was a compound, and its elements had different fates. Proletarianisation, the spread of the wage relation until it becomes the default condition of labour, advanced across the whole planet. Polarisation, the simplification of the class structure into two homogeneous camps, did not occur in the advanced economies in the form envisaged, because the division of labour inside firms and states generated intermediate positions faster than concentration eliminated them, a complication Wright’s contradictory locations were built to absorb.

The honest summary is that the growth prediction succeeded on the relational criterion and the homogeneity prediction failed on every criterion, and that the failure matters. A class that is structurally predominant but internally stratified by credentials, authority, security, sector, and nation does not present the unified face that the older movement expected, and its politics are correspondingly fractured, pulled apart by real differences of interest among its own layers as well as by the cultural and national loyalties that cross-cut class everywhere. Defenders of the persistence case who quote the growth figures without the stratification figures are doing what they accuse the obituarists of doing, reading one row of the table as the whole. The full table supports a more interesting conclusion than either obituary or triumphalism: the working class exists, it is enormous, it is growing, and it is not, in its present composition, the kind of thing that automatically becomes a collective actor. Existence and agency are different properties, and the theory’s confusion of them in its prophetic register is the part of the inheritance that the evidence has genuinely damaged.

What would count as evidence that the working class had really disappeared?

Its disappearance would require the wage relation itself to recede: most people living from property they control, from independent production with secure access to markets, or from provision that does not depend on selling labour power. Falling factory employment, rising affluence, or fading class identity would not suffice, since none touches the relation itself.

That test clarifies the verdict by stating its conditions in advance, which is what the criterion-first rule demands of every participant in this debate. Apply the test to the record and the answer is not close. Property sufficient to live from remains concentrated in a small minority of households in every economy for which distributional data exist. Independent production survives at the margins, much of it dependent. Provision without work, where it exists, is largely derivative of past or family employment. The wage relation, meanwhile, completed its conquest of the world’s labour across the very period under discussion, absorbing the peasantries of Asia, the artisans of the older economies, and profession after profession that once sold its services as an independent calling. Against a disappearance of that magnitude, the disappearance literature offers the decline of a sector, a style, and a self-description. Those declines are facts, and the societies that underwent them were transformed by them, but they are facts of a different order, and the criterion-first rule exists to keep the orders separate.

The recomposition of the class: women, migration, and the new majorities

Any account of the class’s persistence that pictured it as the same population in new jobs would fail the evidence in a second way, because the wage dependent population was recomposed demographically across the same decades. The largest single change was the movement of women into wage employment on a mass scale. In the older economies, the mid-century working class of the classic portrait was disproportionately male, its wage supplemented by unpaid domestic labour that the class analysis of the period largely failed to count. Across the closing decades of the century, married women and mothers entered employment in numbers that transformed both the labour market and the household, until the typical wage dependent household became a two-earner household and the typical employee was as likely to be a woman as a man in many service sectors. Feminist critics inside and outside the Marxist tradition drew the correct conclusion: a theory of class that had treated the male industrial worker as the universal case had been describing a fraction of its own subject, and the unpaid labour of reproducing labour power, cooking, cleaning, bearing and raising the next generation of workers, belonged inside the analysis rather than in its footnotes. That correction enlarges the relational map rather than disturbing it. The care worker, the cleaner, the nurse, and the teacher, occupations feminised in most economies, are wage dependent in exactly the criterion’s sense, and their mass entry into employment is one of the principal mechanisms by which the class grew while the factories shrank.

Migration worked a similar recomposition across national lines. The labour markets of the older economies drew workers from their former colonies, from poorer regions of their own continents, and from across the world, while the new industrial economies drew rural migrants into their factory cities on a scale dwarfing the older movements. The result is that the global working class is more various in language, citizenship, and legal status than any previous formation of wage labour, and its internal divisions are correspondingly real: the migrant without secure residence status sells labour power under a compulsion sharpened by law, the citizen worker beside her under a compulsion softened by rights, and employers have understood the uses of that difference for as long as labour has crossed borders. Class analysis that attends to the relation can register these gradations of dependence precisely, because dependence is its home variable. What the recomposition changes is not the existence of the class but the difficulty of its unity, since the shared relation now has to be recognised across differences of gender, nation, legal status, and language that the older, more homogeneous formations did not have to bridge. The disappearance writers pointed at the fragmentation and saw dissolution. The relational account sees a larger class with a harder organisational problem, which is a different and more useful diagnosis, because it locates the difficulty where remedies would have to operate.

There is a generational dimension too, easily missed because it works slowly. The spread of wage dependence across generations dissolved the mixed household economies in which a wage coexisted with a smallholding, a family shop, or a trade practised independently from the home. Where a grandfather might have combined factory shifts with an allotment and a grandmother’s taking in of work, the grandchildren are typically pure cases: employees whose entire subsistence arrives through the wage, tenants or mortgaged buyers whose shelter depends on it, pensioners whose retirement income is a deferred portion of it. Sociologists call the underlying process the completion of proletarianisation, and it proceeded furthest in exactly the societies whose commentators were most confident that the proletariat had disappeared. The paradox is only apparent. A condition becomes invisible as it becomes universal, the way water is invisible to fish, and a society in which nearly everyone must sell labour power to live may stop remarking on the fact precisely because there is no contrasting condition left to remark against. The cultural criterion then reports, accurately, that class feels weaker than it did, while the relational criterion reports, also accurately, that the relation has never been more complete. Both reports belong in an honest account, and the criterion-first rule is what allows them to be held together without contradiction.

The public sector and the state as employer

A further quarter of the modern wage dependent population raises a question the classic texts handled briefly: the employees of the state. Teachers in public schools, nurses in public hospitals, civil servants, refuse collectors, university staff, and the armed services all sell their labour power, but their employer is not a capitalist seeking profit from their labour, and a purist version of the theory hesitated over whether they belonged to the working class at all or to some salaried estate outside the capital relation. The hesitation generated a genuine internal debate, tangled with the separate question of productive and unproductive labour, and the disappearance writers occasionally borrowed it, suggesting that the growth of public employment marked an exit from class society into a administered order of functionaries.

The relational criterion, applied with the same operative test as elsewhere, gives a steady answer. The state employee owns no means of production, lives by a salary, works under direction within a hierarchy, and is exposed to dismissal, reorganisation, and the fiscal weather of governments. Her employer’s purposes differ from a firm’s, and the surplus question takes a different form, since her salary is paid from taxation rather than advanced as capital, but her position in the structure of dependence is the wage dependent position, and the history of public sector organisation, strikes, and pay disputes shows the position generating the same repertoire of conflict as private employment whenever its terms are pressed. What public employment changes is the political geometry around the relation: the employer is ultimately answerable through elections, the service’s users are the worker’s fellow citizens, and disputes are fought in the open as arguments about taxation and provision. Those differences matter enormously for strategy and for public sympathy. They do not relocate the nurse or the teacher outside the class that the criterion defines, and the growth of public employment across the twentieth century counts, on the relational ledger, as another of the great channels through which wage dependence became the normal condition of the population.

Identity, voting, and the evidence about consciousness

Since the concession about consciousness carries so much of the objection’s weight, the evidence deserves a section of its own, stated as even-handedly as the debate allows. Begin with self-identification. Survey researchers who ask people to place themselves in a class find, across many countries and decades, a persistent pattern: large numbers continue to describe themselves as working class, including many in occupations the occupational map calls middle class, while the share choosing working class labels has declined in several countries and the share refusing class labels altogether has grown. The pattern resists both camps’ slogans. Identity has not vanished, and it has not held its old form. It survives as a residue of family origin, a description of where a person started and what their parents did, as much as a reading of their present position, which is why graduates in salaried professions may claim a working class identity their incomes contradict while manual workers with businesses claim none. The cultural criterion, to its credit, predicted this kind of untidiness, since identities are inherited and narrated rather than recalculated from payslips. What the surveys cannot show, on any reading, is the one thing they are routinely cited for: the non-existence of a structural class. They measure the shadow the structure casts on self-understanding, and a fading shadow is a fact about light and angle as much as about the object.

Voting tells a parallel story with the same moral. The correlation between occupational class and party choice, strong in the mid-twentieth century in many democracies, weakened markedly across its closing decades, as the great class parties broadened their appeals, as new cleavages of education, age, nation, and values cross-cut the old one, and as turnout itself fell unevenly across the classes. Political scientists who announced the end of class voting and those who replied that class voting had changed its form rather than ended it were, on inspection, replaying the criterion problem in miniature, since the finding depended on how classes were coded and which elections were chosen. The substantial point survives the technical quarrel: class position became a weaker predictor of the vote, full stop, and any theory that had staked itself on a tight correspondence between position and party has to absorb that weakening as a loss. What does not follow is the inference the obituaries drew, from weaker class voting to a vanished working class. Voters did not cease to be wage dependent when they ceased to vote as their grandparents had. They ceased, in large numbers, to find a political vehicle that organised their dependence into a programme, which is a fact about parties and movements, the supply side of politics, as much as about the electorate’s structure.

Organisation completes the triptych. Union membership density fell across most advanced economies from its late twentieth-century heights, driven by the sectoral shifts already described, by legal and political defeats, by employer resistance, and by the growth of employment forms, small workplaces, dispersed services, casual contracts, that are hard to organise by the methods the industrial unions perfected. The fall is the single most concrete institutional fact behind the disappearance literature, and its consequences are not in dispute: where density fell, the wage share, workplace protections, and the everyday experience of collective presence fell with it. Yet density is a variable with a history of rising as well as falling, and it rose, in its time, among exactly the service and public workforces whose growth was supposed to make organisation obsolete. The relational reading treats organisation as a contingent achievement built on a structural base, sometimes present and sometimes absent, and therefore treats its decline as a change in the class’s condition and power, grave in itself, but categorically different from the class’s abolition. A army that has lost a war has not been shown never to have existed, and a class that has lost organisational ground has not been shown to have dissolved into the population at large. The distinction sounds elementary. The literature of disappearance depends on blurring it.

Why the obituary keeps being written

It is worth asking, finally, why the announcement of the working class’s death recurs with such regularity, in new vocabularies, across more than a century. Versions of it predate the post-industrial thesis by generations: the embourgeoisement claims of the late nineteenth century, the managerial revolution arguments of the mid-twentieth, the affluence arguments of the postwar boom, and the knowledge economy arguments of the digital era all announce, in their period’s idiom, that the classic opposition of capital and labour has been overtaken. The recurrence itself is data. A finding that has to be rediscovered every thirty years, against a world that declined to notice the previous discoveries, suggests a genre at work rather than a measurement improving. The genre has constant temptations. It flatters the present as an emancipation from the crude conflicts of the past. It relieves parties of the embarrassment of their old electorates and workers of the unfashionable identity itself. It mistakes, each time, the most visible contemporary transformation, affluence, then services, then information, then platforms, for the abolition of the relation that persists through all of them by changing its clothes.

None of this makes the genre worthless. Each wave of obituary has identified a real transformation that class analysis needed to absorb: the genuine softening of material extremes in the affluent decades, the genuine centrality of knowledge and credentials in the service economy, the genuine fragmentation of the platform workplace. The tradition is healthiest when it treats the obituaries as its research agenda rather than as its refutation, extracting from each the change it must explain and discarding the headline it cannot support. That is, in effect, what the strongest work in the tradition did. Wright built contradictory locations out of the managerial objection and credential assets out of the educational one. The precariat debate forced a reckoning with security as an internal division. The global relocation argument forced the map’s scale from the national to the planetary, where the relational criterion’s results are least ambiguous. A theory that grows by digesting its obituaries is not performing the auxiliary hypothesis manoeuvre the objection alleged. It is doing what living theories do, provided the core criterion stays fixed while the elaborations multiply, and the criterion, ownership against the need to sell labour power, has stayed fixed since the beginning. The obituary writers, for their part, perform a service they did not intend: they keep the criterion question in public view, and every few decades a new audience gets to learn that the answer depends on what is being counted, which is the first thing this debate should teach.

The income criterion examined: what gradients can and cannot show

Because the income criterion is the one most often substituted for the relational map in public argument, it merits a direct audit of its own. Its appeal is obvious. Income is measurable, continuous, and available in long statistical runs, and a society sorted by income into bands offers the comforting picture of a bulging middle with thinning extremes, a diamond replacing the pyramid of the industrial age. From that picture the conclusion is drawn that classes, understood as bounded and opposed groups, have been replaced by a gradient along which individuals move up and down across a lifetime. The picture is not false. The conclusion does not follow, and examining why it does not follow teaches more about criteria than any other comparison in this debate.

An income band is a slice of a distribution, and its boundaries are chosen by the analyst. Set the middle band generously and almost everyone is middle class; set it narrowly and the middle evaporates. The measured size of the working class under an income criterion is therefore partly an artefact of the ruler, and rival studies using rival rulers duly report rival middles, a warning sign that the instrument is recording its own settings. Worse, income conflates flows from opposite relations. A skilled employee’s high salary and a small proprietor’s moderate takings may place them in the same band while their sources, security, and exposure differ in kind. The employee’s income stops when the employment stops and is set within an organisation’s hierarchy. The proprietor’s continues while the asset works and answers to the market alone. Conversely, the income map separates people the relational map unites, placing the senior nurse and the junior nurse in different classes although they sell their labour power to the same employer on the same corridor, and unites the nurse with a landlord of similar income although his living arrives whether he rises in the morning or not. A criterion that cannot distinguish living from labour from living from property has given up the distinction the whole concept of class was invented to carry.

The gradient picture also mishandles time. Movement along an income distribution across a lifetime, the young worker poor, the established worker comfortable, the pensioner modest again, is real, and it softens the felt rigidity of class. But a life cycle pattern is not a class structure dissolving; it is individuals traversing positions that persist independently of who occupies them. The low paid jobs of youth are refilled by the next cohort as reliably as they are vacated, and the question the relational criterion asks, what positions exist and how people are compelled into them, is untouched by the observation that particular people pass through several positions in a life. Mobility studies, the income tradition’s most sophisticated branch, in fact confirm the structural point they are sometimes enlisted against: measured across generations, movement between the children of wage earners and the children of owners remains limited and asymmetric, with the ownership position in particular reproduced through inheritance with a tenacity no income band captures, since the inherited asset may yield little income in the years it is quietly appreciating. Wealth, the stock, tells the story the flow conceals, and every serious distributional study finds the stock far more concentrated than the flow, with the greater part of productive wealth held by a small fraction of households even in the most egalitarian of the advanced economies. Read by wealth rather than income, the society of the bulging middle shows a familiar outline: a small stratum whose holdings could sustain it indefinitely, and a vast remainder for whom the wage is not a preference but the condition of the month’s bills. That outline is the relational map, arriving by the income tradition’s own road once the right variable is chosen.

None of this makes income analysis useless, and the criterion comparison table credited it with its real virtues: it tracks living standards, poverty, and the material texture of inequality better than any rival, and policy made without it would be blind. The claim here is jurisdictional. Income measures how much people have. Class, in the sense the existence debate inherited from Marx, names how they are placed in the organisation of production, and the two questions intersect without merging. Most confusion in the debate can be traced to a jurisdictional error, an income finding announced in class vocabulary, and the criterion-first rule is simply the demand that the announcement name its jurisdiction before its conclusion.

The new industrial world: the class where the factories went

Return to the global dimension once more, because it is the part of the argument most often nodded at and least often absorbed, and because a debate conducted entirely within the older economies systematically distorts it. When manufacturing employment contracted in the North Atlantic world, the standard narrative at home spoke of an economy passing beyond industry into services and knowledge. From the standpoint of the world economy, no such passage occurred. Industrial production grew across the period, enormously, and its geography shifted toward East and South Asia, where new industrial workforces were assembled in a few decades on a scale that took the older economies a century to reach. The workers of the new industrial regions were recruited overwhelmingly from the countryside, from peasant households whose relation to the land was dissolved by the same mixture of opportunity and pressure that had filled Manchester and the Ruhr in their centuries. They entered factories, dormitories, and construction sites as sellers of labour power in the purest form the criterion recognises, separated from the means of production their families had worked for generations and dependent on the wage for every necessity of an urban life. Whatever else the late twentieth century was, on the relational map it was the greatest single expansion of the working class in history, concentrated in the decades when the class’s death was being announced in the countries it was leaving.

The conditions of the new industrial world also supply a control experiment for the disappearance thesis’s favourite variables. The new workforces were, by the standards of the older economies at the time, low paid, weakly protected, and housed in conditions of a starkness that recalled the early industrial city. As industrialisation proceeded, incomes rose, consumption spread, workers acquired the telephones, motorcycles, and eventually the homes and pension entitlements whose acquisition in the older economies had been cited as proof that workers were ceasing to be workers. The relational prediction is that none of this acquisition moves the class boundary, because the boundary was never drawn at a standard of living. The record conforms. Prosperous industrial workers in the new economies remained wage dependent, organised where they could, struck where they judged it necessary, and discovered in downturns the same exposure their counterparts elsewhere know. Affluence arrived, as it had in the older economies, as a change in the class’s circumstances rather than as its dissolution, and the embourgeoisement hypothesis failed its second trial on a second continent for the same reason it had failed its first: it mistook the contents of the worker’s home for the source of the worker’s living.

The global frame also corrects a subtler parochialism, the assumption that the older economies’ occupational structure is the terminus of development rather than a local configuration. The service heavy economies of the North Atlantic rest, in material terms, on industrial production carried out elsewhere, and their wage dependent populations are linked to the new industrial workforces through the supply chains of the same firms, a single division of labour spanning the warehouse picker in one country and the assembler in another. A class analysis confined to one nation’s statistics will describe the picker as a service worker in a post-industrial society and the assembler as a member of a rising industrial class abroad, and will miss that they occupy connected positions in one structure, both selling labour power to organisations owned in a third place. The relational criterion, indifferent to the statistical border, sees the structure whole. That indifference is among its strongest claims to be the right instrument for the existence question in a globalised economy, and among the clearest reasons the obituaries, written nation by nation with occupational rulers, kept arriving at a death the world economy had not suffered.

Three households through six criteria

Abstract criteria earn their authority in application, so consider three households and watch the maps disagree about them in instructive ways. The first household is a senior hospital nurse and her partner, a secondary school teacher, in a mortgaged house in a northern English city, with two children, modest pension entitlements building on both salaries, a car each, and holidays abroad in most years. The second is a self-employed builder in the same city, owning his van, his tools, and a reputation that keeps his order book full, employing an apprentice in busy seasons, earning in a good year roughly what the nurse earns and in a bad year a good deal less, renting the house he lives in. The third is a warehouse picker and her partner, a cleaner employed by a contracting firm, in social housing across the same city, with earnings between them below the builder’s good year, no savings beyond a small buffer, and pension entitlements that are recent and thin.

Run the households past the criteria. By income, the nurse and teacher stand a full band above the picker and cleaner, with the builder oscillating between the two bands by the year, so the income map draws its class line somewhere through the builder’s fluctuating takings and pronounces the first household comfortably middle class. By manual work, the builder and the picker are working class together, the teacher and nurse are not, and the cleaner is working class only if cleaning counts as manual, a classification quarrel that has consumed real scholarly time. By cultural identification, answers scatter with biography: the nurse, a miner’s daughter, may insist on the identity her salary contradicts, the builder may reject any label that groups him with employees at all, and the picker may use the term without hesitation. By property in the relational sense, the map returns its characteristic and, to many eyes, perverse verdict: the nurse, the teacher, the picker, and the cleaner are in one class, all four living wholly by the sale of their labour power, while the builder, poorer than the nurse and rougher in every cultural marker, stands outside it as a small proprietor, because his van, tools, goodwill, and order book are means of production that he owns and works, and because a month without an employer’s permission is his normal condition rather than his catastrophe.

The verdict is the strongest objection in concrete form, and this article has already conceded what is true in the objection: the line so drawn does not track income, comfort, culture, or felt solidarity, and anyone who wants a map of those things should use a ruler built for them. What the line tracks is a structure of dependence and its characteristic emergencies. Let the hospital and the school close in the same season, an event no more fantastic than the pit closures the region remembers, and the first household discovers its class position in a quarter: two salaries gone, a mortgage running, pensions frozen in place, and nothing to sell but labour power in a labour market suddenly crowded with sellers. Let the builder’s market collapse and he discovers a different emergency: no dismissal, no redundancy payment, no employer at all, only the silence of the telephone and the slow consumption of his capital’s value, an emergency of the market rather than of authority. Let the warehouse automate and the third household meets the first household’s emergency without the buffer of savings or the cushion of scarce credentials, which is a difference of degree inside a shared structure, not a difference of structure. Emergencies are the moments when class positions become visible, because they strip away the mediations, the payslip’s size, the job’s prestige, the home’s comforts, that disguise them in ordinary times, and a criterion that groups people by the emergency they would face is grouping them by something real, even when the grouping offends the eye in the calm.

The households also show why no single alternative map can simply replace the relational one in the existence debate. Suppose the income map were adopted as the official meaning of class. Then the working class’s existence would become a question about the shape of the earnings distribution, and its alleged disappearance would mean only that earnings had compressed or risen, a claim about prosperity wearing the vocabulary of structure. Suppose the cultural map were adopted. Then existence would turn on self-description, and a successful campaign of relabelling, a generation taught to call itself middle class, would abolish the working class by nomenclature while every relation in the economy stood unchanged. The disappearance conclusions that animate the debate derive their drama from the relational meaning of class, the meaning in which a class’s end would be a world historical event, while their evidence is gathered on maps where the conclusion means something far smaller. Holding the two apart, the drama from the evidence, is the whole of the criterion-first rule, and the three households are its demonstration in miniature.

The knowledge economy version of the argument

Each era’s disappearance thesis speaks its era’s language, and the version native to the digital economy deserves separate handling because it is the form in which many readers will have met the argument. It runs as follows. Production increasingly consists of information, design, and code rather than of material goods. The decisive asset of the new economy is knowledge, and knowledge, unlike land or machinery, is carried in the heads of the workforce, who therefore own the true means of production of the age. The knowledge worker is a proprietor in disguise, renting out an asset she can take from firm to firm, increasingly able to work from anywhere, for anyone, or for herself, and the old drama of capital confronting labour dissolves into a market of talents contracting with platforms and firms on something like equal terms. The industrial categories, on this account, were not refined by the service transition; they were retired by it.

The argument identifies a real shift in production’s texture and then draws from it a conclusion the relational criterion blocks at three points. The first concerns what knowledge workers actually own. A skill carried in the head is embodied in a person who must present herself, in person or online, for hire, hour by hour, and who cannot detach the asset, lend it out, or live from its yield in absence. The earlier discussion of credential assets established the general point: embodied assets are collected only through the sale of labour power, and their rents, where scarcity grants them, are a wage premium rather than a property income. The knowledge economy multiplies such assets without altering their nature. The second point concerns the means that knowledge production actually requires. Code is written on rented cloud infrastructure, distributed through platforms and application marketplaces owned by others, financed by capital raised from owners, and sold inside ecosystems whose owners set the terms of access and take their share at the gate. The programmer’s laptop is the builder’s van at a higher level of abstraction: a genuine tool, genuinely hers, and wholly insufficient as a means of production when the roads, the market, and the customers belong to someone else. The third point is empirical in the plainest sense. The knowledge industries are staffed overwhelmingly by employees, salaried, managed, monitored by metrics of a granularity the factory never achieved, and dismissed in waves when their employers’ valuations turn, and the celebrated independence of the star practitioner, like the star athlete’s, is a property of scarcity at the summit rather than of the occupation’s structure. A category that fits the summit and not the structure is a biography, not a class analysis.

Where the knowledge argument does bite, it bites in a place the relational tradition should acknowledge cleanly: the dispersion of tiny equity holdings and the blurring of the wage at the top of the technology sector, where compensation in shares can, for a minority, accumulate into genuine property within a working lifetime. The tradition’s machinery handles the case by proportion, as it handles the senior executive elsewhere, classifying by the dominant source of a household’s living. What the case does not do is generalise downward. For every employee whose shares matured into independence, the sector employs multitudes for whom equity was a lottery ticket stapled to a salary, and the salary remained the substance. The knowledge economy, audited by the operative test, turns out to have reproduced the wage relation inside the newest industries with the same completeness as the service economy before them, adding its own dialects of direction, the dashboard, the sprint, the review cycle, to the long vocabulary of managed labour. The costume changed again. The relation, once more, did not, and a reader trained by the criterion-first rule to separate costume from relation will find the newest obituary the easiest of all to read.

Status, consumption, and the death of class thesis in depth

The most theoretically ambitious of the disappearance arguments, the death of class thesis of the mid-1990s, deserves a slower hearing than the summary given earlier, because it does not rest on factory counts at all and cannot be answered with them. Pakulski and Waters argued that the advanced societies were moving from a class regime to a status based order. In the class regime, position in production governed life chances, identity, and conflict. In the emerging order, they proposed, stratification is increasingly organised around consumption, lifestyle, cultural capital, and status honour, distributed through markets and media rather than through the workplace, so that the great collective actors dissolve into audiences and taste publics. The thesis had genuine explanatory targets in its sights. The politics of the period did increasingly organise around cultural and status questions. Consumption had become a principal theatre of identity, with brands and lifestyles doing signalling work that occupation once did. Educational credentials, on their account a status good as much as a productive asset, sorted life chances with a finesse the crude ownership line could not match. If the twentieth century’s master division was between those who owned production and those who worked, the thesis asked whether the twenty-first’s might be between the credentialed, the connected, and the culturally fluent on one side and the excluded on the other, a division that cross-cuts the wage relation so thoroughly that class becomes a legacy vocabulary.

The reply begins by granting the phenomenon and disputing the succession. Status competition and consumption signalling are not new orders replacing an older one; they are permanent features of stratified societies that class analysis always coexisted with, visible in the sumptuary anxieties of the nineteenth century as plainly as in the brand economies of the present. The question is whether the new currencies float free of the old structure or are minted within it. Consumption capacity is, for the overwhelming majority, a function of the wage and its security, and the lifestyles that status markets sell are priced against salaries with actuarial precision. Cultural capital converts, reliably and at published rates, into educational credentials, and credentials into positions inside organisations owned by others, which is to say that the status order operates as a sorting mechanism for the class structure rather than as its replacement: it decides which wage dependent person gets the secure, autonomous, well paid post and which gets the monitored and precarious one. That is a profound function, and an analysis that ignores it will misunderstand how the class is internally divided and why its members so often experience one another as strangers. But a sorting mechanism inside a structure is not a new structure, and when the economy contracts, the dependence beneath the sorting reasserts itself in the order the relational map predicts: the precarious first, the monitored next, the credentialed last and least, all of them sellers of labour power discovering the terms of their sale in sequence. The death of class thesis mistook a change in how class positions are allocated and experienced for the death of the positions themselves, which is the criterion error in its most sophisticated dress, a cultural instrument’s reading announced as a structural finding.

A similar depth test can be applied to individualisation, the other sophisticated pillar of the disappearance case. Beck’s argument was that the institutions of industrial society increasingly address people as individuals, that welfare, law, and the labour market all require biographical self-management, and that class cultures consequently thin into private fates. The observation is acute, and any honest persistence case must absorb it: the experience of class has been individualised, its risks privatised, its setbacks narrated as personal failures in the idioms of therapy and self-improvement. But individualised experience is not individualised structure. The redundant worker who blames herself is redundant by a decision taken in a structure she shares with thousands, and the fact that she narrates the event privately measures the weakness of the collective vocabularies available to her, not the absence of the collective condition. Here again the two questions this article separated at the outset, position and consciousness, do all the work. Individualisation is a thesis about consciousness and its institutional supports. Read as such, it is largely persuasive and entirely compatible with the persistence of the class it is said to dissolve. Read as a thesis about position, it asserts that because people face the labour market one by one, the labour market has ceased to be a structure, which does not follow and which its author’s own risk analysis, with its insistence that risks are socially produced and unequally distributed, implicitly denies. The sophisticated disappearance arguments, examined closely, repeatedly turn out to be strong theses about identity, culture, and organisation wearing the titles of theses about existence. The criterion-first rule does not refute them. It re-shelves them, on the shelf where they are valuable.

Retirement, the deferred wage, and the class across a lifetime

One more region of the map repays attention, because a large and growing share of the population stands in it: retirement. The pensioner neither sells labour power nor, in the main, owns means of production sufficient to live from. Does she exit the working class at the pension’s first payment, and if the retired are counted out, does the class shrink by exactly the population that its own history produced? The question sounds like a scholastic puzzle, and it opens onto the substance of what a pension is. In the contributory systems of the older economies, the pension is best understood as a deferred wage, a portion of the lifetime price of labour power withheld and returned across the years when labour power can no longer be sold. Its level is set by the history of the wage, its security by the institutions the labour movement built, and its adequacy by the same distributional contests that govern pay. The pensioner’s material life is the wage relation extended across the life course, and class analysis has generally counted the retired by the position from which they retired, a convention that the income criterion, with its habit of filing pensioners as a separate low income stratum, persistently mangles. On the income map, retirement looks like a descent into the working class or out of the middle, depending on the bands. On the relational map, nothing structural happens on the birthday in question: a lifetime’s dependence on the wage completes itself in dependence on the wage’s deferred portion, and the vulnerabilities persist in translated form, exposure to inflation, to pension fund solvency, to the fiscal politics of governments courting or neglecting the retired vote.

The lifetime perspective generalises. Class position, as the relational criterion defines it, is most accurately predicated not of a moment but of a trajectory, and much apparent movement between classes is the life course passing a single person through the positions in sequence: the student with a service job, the graduate employee, the manager in contradictory location, the pensioner. Snapshots mislead in both directions, flattering the dissolution case with the comfort of the established years and flattering nobody with the insecurity of the starting ones. The snapshot errors compound in cross-section statistics, which photograph the whole population at an instant and sort the young, the established, and the retired into different bands as though they were different populations. A debate conducted in snapshots will keep rediscovering that the working class is simultaneously shrinking, since the established are comfortable, and growing, since the young are precarious, and will mistake its own method for a paradox in the world. The criterion-first rule, applied to time, instructs the reader to ask of any class statistic whether it describes positions or the momentary occupants of positions, and to expect the existence question to be answerable only in the vocabulary of positions. In that vocabulary, the retired nurse belongs to the history and the present of the working class, her pension a claim on the social product earned in the same relation her grandchildren are entering at the warehouse and the hospital, and the intergenerational household, in which wages and pensions pool to meet a single rent, is the class’s most ordinary economic unit. Obituaries that count only the employed, like maps that count only the manual, arrive at their deaths by the quiet choice of what to leave out.

The professions: how independent callings became salaried posts

The professions supply the persistence case with its most quietly decisive evidence, because their transformation is documented in the biographies of institutions rather than in contested statistics. Consider the trajectory of medicine. The doctor of the classic portrait was an independent practitioner, owning a practice, selling services to patients in a market of reputations, and standing in the old taxonomy among the petty bourgeois or the liberal professions outside the wage relation altogether. The doctor of the contemporary hospital is a salaried employee, working shifts set by an organisation, practising under protocols and budgets she does not control, her clinical autonomy real but bounded by the institution that buys her labour power. Law followed a version of the same path: the independent advocate persists, but the mass of legal labour is performed by salaried solicitors, employed counsel, and associates inside firms whose ownership sits with a narrow partnership tier, and below them armies of paralegals and clerks whose wage dependence nobody disputes. Academia’s version is the more pointed for being recent in its acceleration: the scholar as independent man of letters gave way to the salaried professor, and the salaried professor is giving way, at the margin where the system’s growth occurs, to the adjunct and the fixed term teaching fellow, whose condition, doctorate notwithstanding, is precarious wage labour in its textbook form, hired by the course, paid by the term, and separated from the means of scholarly production, the library, the laboratory, the press of reputation, unless an institution grants access.

Engineering, accountancy, architecture, journalism, and the clergy each tell a local variant of the story, independent practice contracting into salaried employment within organisations owned or governed by others. The significance for the existence debate is that these were the occupations the two-class simplification was always accused of ignoring, the standing exhibit for the claim that modern societies generate a third thing, neither capital nor labour, that dissolves the old polarity. The exhibit has changed its testimony. What the professions demonstrate, on inspection, is not a third relation but the wage relation’s capacity to absorb the highest credentialed labour on the same structural terms as the lowest, modulated by scarcity rents, professional closure, and collegial self-government, all of which soften the experience of dependence without abolishing its structure. The professional’s day feels nothing like the picker’s, and this article’s concession about the category’s breadth stands. But when the hospital trust restructures, the firm merges, or the university cuts, the professional meets the same instruments as every other employee: the consultation document, the redundancy round, the frozen establishment, the discovered truth that the salary was the substance and the status the wrapping. A criterion that predicted that meeting, and located these occupations inside the relation before the meeting occurred, has earned something the obituary tradition never credits it with: it saw the direction of travel of the most prestigious labour in the society, and it saw it early, when the independent practitioner was still the professions’ self-image. The growth of the working class on the relational criterion is not a statistical trick performed with low paid multitudes. It is the recorded fate of occupation after occupation that once stood, or believed it stood, outside the sale of labour power altogether.

Why both sides can cite the same street

The whole debate can be walked along a single ordinary street in a deindustrialised town, and the exercise is worth performing because it explains the argument’s emotional durability better than any statistic. At one end stands the converted mill, its floors now apartments, its chimney preserved as heritage. A disappearance writer reads the building correctly as an ending: the hundreds who once worked there as a single workforce, in a single relation to a single owner, are gone, and with them the union lodge, the brass band, the occupational community that made class a lived identity rather than a sociological term. Halfway along the street, the persistence writer points to what replaced the mill’s economy: a distribution depot on the ring road employing more people than the mill did, a hospital expanded into the town’s largest employer, a retail park, a further education college, care agencies sending workers into homes on every street, and the converted mill’s own residents leaving each morning for salaried jobs in offices two towns away. Every one of those workers sells labour power. Almost none of them experiences the sale as membership of a class with a name, a headquarters, and a history. Both writers are describing the street accurately. They are describing different properties of it, solidity of organisation in the one case, structure of dependence in the other, and the entire literature of disappearance and persistence is contained in the distance between those two descriptions. The criterion-first rule does not ask either writer to stop describing. It asks each to label the description with the property it measures, after which the street stops being a paradox and becomes what it is: a place where a class lost its classical form and kept its defining relation, in full view of everyone, argued about ever since.

The street also teaches the debate’s final methodological lesson, which generalises far beyond this question. Whenever a large social category is pronounced dead, the first investigative step is not to gather fresher evidence but to interrogate the instrument that produced the death certificate. Categories die in two ways: because the world stops containing what they name, and because the observer swaps the instrument that detects them for one that detects something else. The two deaths look identical in a headline and are opposites in substance, since the first is a discovery and the second is an error. Class is the standing example because its instruments are so easily confused, hands for relations, incomes for positions, identities for structures. But the discipline the criterion-first rule imposes, name the test before the result, applies wherever a contested category is counted, and a reader who carries it out of this debate into every other argument about vanishing and emerging groups will find that a surprising share of social controversy dissolves, on contact, into instruments talking past one another.

The verdict

The debate, reconstructed, resolves into three findings. The first is that the question as usually posed is malformed, because it invites an empirical answer to a definitional dispute, and the criterion-first rule is the correction: state the criterion, and the answer is arithmetic. On the manual, income, and cultural criteria the working class of the older economies has shrunk, in some respects dramatically. On the relational criterion it has grown there and grown enormously worldwide. The second finding is that the criteria are not equal in standing, because they answer to different purposes, and the relational criterion is the one that the theory of class, as a theory of production and power, is actually about. Its rivals measure the class’s costume, its income, and its self-image, all worth measuring, none identical to the class. The disappearance thesis, in its sophisticated forms, is therefore best read as a true report filed under a misleading headline: the industrial form of the working class, and much of its classical culture and organisation, have passed away in the older economies, while the relation that defined the class all along extended itself over the greater part of humanity. The third finding carries the concession that keeps the verdict honest. The theory’s expectations about consciousness and agency, the passage from a class in itself to a class for itself, have not been vindicated, and the distance between the class’s structural weight and its political presence is the outstanding unsolved problem the debate leaves behind. It is a problem about organisation, culture, and institutions. It is not a problem about existence, and treating it as one has wasted several decades of argument that the criterion-first rule would have saved.

For readers who keep their working texts close, the definitional machinery of this debate repays the same care as any set of accounts: state the criterion before the count, keep the six criteria separate in the notebook, and record which map each new claim about class is drawn on. The study notes collected at VaultBook are built for exactly that discipline, and the criterion comparison table above transfers to a notes page without alteration, which is more than can be said for most of the claims this debate produces.

For students and exam candidates

This debate is a gift to the prepared candidate, because most answers fail it in the same way: by marshalling evidence for one side without noticing that the question turns on a definition. Build the essay on the criterion comparison table. Open by stating the criterion-first rule and naming the criteria in play, manual, income, cultural, and relational. Then give the disappearance case its strongest evidence, factory decline, embourgeoisement, fading identity and organisation, before showing which criterion each piece of evidence is measured on. Introduce Wright’s contradictory locations and the credential asset to handle managers and graduates, since examiners reward the candidate who can complicate the map without abandoning it. Close with the defended verdict: existence on the relational criterion, decline on the occupational and cultural ones, and the consciousness gap conceded as the theory’s genuine unsolved problem. An answer structured that way demonstrates the skill the question is really testing, which is the separation of a definitional dispute from an empirical one.

For teachers and seminar leaders

Run this debate as a discipline exercise before it becomes a contest of loyalties. Divide the seminar and require each side, before any evidence is admitted, to state its criterion in one sentence and to specify what observation would count against its own position. The disappearance side must say whether factory counts, incomes, or identities carry its claim; the persistence side must say what a world without a working class would actually look like. Only then release the evidence, and let students discover that most exhibits support both sides once the criteria are separated, which is the lesson the session exists to teach. The three households exercise works well in the room: give the groups the nurse and teacher, the builder, and the picker and cleaner, and have each group classify all three under an assigned criterion before the maps are compared. Close by asking which criterion a class theory is for, and let the argument about purposes, rather than about counts, occupy the final minutes.

For researchers, historians, and journalists

The working class is declared dead in print with a regularity that should embarrass the declarers, and this article’s machinery is designed for the desk where those declarations are drafted. Before publishing any version of the claim, run three checks. Identify the criterion the piece is using, and if the piece cannot identify one, that is the finding. Separate the occupational evidence from the relational evidence, and refuse to let a manufacturing series carry a conclusion about wage dependence. Ask whether the claim is national while its subject is global, since the class’s centre of gravity moved while the obituaries were being written in the older economies. Researchers will find the live questions in the operationalisation of the relational criterion at its edges, dependent self-employment, platform work, and the coding of authority and credentials in survey data, where Wright’s programme remains the benchmark. Historians should note how completely the disappearance narrative depends on taking the mid-century industrial formation as the class’s essence rather than as one of its configurations. The story is never that nothing changed. It is that the change has a name more precise than death.

Frequently Asked Questions

Q: Does the working class still exist?

On the criterion Marx’s theory uses, yes. Class there is defined by relation to the means of production, and the working class consists of those who, lacking productive property sufficient to live from, must sell their labour power to an employer. By that test the class includes manual, service, clerical, technical, and professional employees alike, and it has grown as wage dependence has spread across sectors and continents. What has shrunk, sharply, is the industrial and manual form the class took in the older economies, together with much of its classical organisation and identity. Whether the answer is yes or no therefore depends on the criterion smuggled into the question, which is why this article insists on the criterion-first rule: state the test, and the existence question answers itself; leave the test unstated, and the argument can continue forever without the parties ever counting the same thing.

Q: Is everyone middle class now?

Only on criteria that define the middle by income, consumption, or aspiration, and those criteria can be tuned to produce almost any size of middle the analyst prefers. Rising living standards, mass home ownership, and expanded education have certainly made the circumstances of wage earners more various and, for many, more comfortable than the industrial past. But the relational question is untouched by comfort. A household lives in the middle of the income distribution and remains working class in structure if its living depends entirely on selling labour power, as the first redundancy notice demonstrates. The feeling that everyone is middle class also draws on the shrinking of the extremes’ visibility: the genuinely property owning stratum is small and socially distant, while the destitute are marginalised out of view. Between them stands a vast population of employees whose shared dependence is disguised by wide differences of salary and style, which is a fact about the class’s internal stratification, not about its replacement by a universal middle.

Q: Does owning a home stop you being working class?

No, because the operative test is not whether a person owns an asset but whether the asset frees them from selling their labour power, and an owner occupied home does not. The house yields shelter rather than income, can be turned into money only by giving up living in it or borrowing against it, and where it is mortgaged it binds its owner to the wage more tightly than renting does, since the payment continues whether the job does or not. Home ownership changes a worker’s life profoundly: it builds a store of wealth, softens old age, and creates an interest in property values that can pull politics in new directions. Those effects are real and class analysis should register them. They alter the experience and the interests of the class’s members. They do not alter the relation that defines membership, which is why the mortgaged nurse and the renting nurse occupy the same position on the relational map while differing on every map of wealth and security.

Q: Are office and service employees working class?

On the relational criterion, yes, whenever they live by selling their labour power, which the great majority do. The criterion sorts by relation to the means of production rather than by the content of the work, so the clerk, the call centre agent, the retail assistant, the care worker, and the software developer employed by a firm all stand in the same structural position as the factory hand: propertyless in the operative sense, directed by an employer, and exposed to dismissal and reorganisation. Older usage restricted the term to manual workers, and under that manual criterion office and service staff fall outside by definition, which is one reason the existence debate talks past itself. Within the relational answer there is room for every real distinction of pay, skill, authority, and prestige among these employees, and Wright’s framework of contradictory locations and credential assets exists to map exactly those distinctions. Inclusion in the class is the beginning of the analysis, not a claim that all its members’ lives are alike.

Q: Has the working class shrunk or grown globally?

Grown, and by a very wide margin, on the relational criterion. Across the period in which manufacturing employment contracted in the older industrial economies, industrialisation in East and South Asia and urbanisation across the developing world drew hundreds of millions from peasant agriculture and informal subsistence into factories, construction, transport, and urban services. Each such movement is a movement into wage dependence, the condition the criterion counts. Even within the older economies, the employee share of the working population rose as self-employment and independent practice contracted. The shrinkage is real but specific: it belongs to the manual and industrial segment inside the older economies, and to the class’s share of employment in particular nations whose production moved abroad. A global count on a relational test therefore reverses the national obituary. The working class did not contract across the late twentieth century. It changed its location, its sectors, and its composition, and became, for the first time, a genuinely worldwide majority condition of labour.

Q: Is the precariat a new class?

The precariat thesis, associated with Guy Standing, proposes a new class in the making defined by insecure work, unstable occupational identity, and exclusion from the employment based securities of the postwar settlement. The standard Marxist reply is that this describes a condition within the working class rather than a new class, because class in the relational sense turns on position in the relations of production, and the precarious worker’s position, propertyless and dependent on selling labour power, is the classical one in an exposed form. Insecurity, moreover, is historically the older condition: casual hiring was the norm for much of the nineteenth-century working class, and the secure lifetime job was the exception built by twentieth-century organisation. The thesis nonetheless registers something real, the widening internal division of the wage dependent population by security and status, and the distinctive resentments that division breeds. This article’s verdict is that precarity grades the working class internally and transforms its politics, without founding a new class on a new relation to production.

Q: Do pensions make workers part owners of capital?

They make workers holders of claims on capital, which is a different thing from owners of it in the sense class analysis uses. A pension entitlement is a right to a future income stream, funded by investments the contributor does not choose, in firms she does not direct, under managers she cannot appoint or remove. The decisive strand of ownership, control over the asset and over the labour it employs, is entirely absent, and the capital assembled from workers’ deferred wages is in practice deployed by fund managers under the same return discipline as any other capital, at times against the employment interests of the contributors themselves. The operative test settles the membership question as it settles the housing question: the employee who stops working years before pension age cannot live on the fund, and must sell her labour power or go without. Pension wealth changes workers’ security and their stake in financial stability. It does not relocate them across the line between those who live by property and those who live by labour.

Q: What is a contradictory class location?

The term is Erik Olin Wright’s, from his reconstruction of Marxist class analysis in the late 1970s, and it names a position that shares the defining features of two classes at once. Managers are the standard case: as salaried employees they sell their labour power and are subject to owners above them, yet they exercise the owner’s functions of directing, supervising, and disciplining other workers, so they stand simultaneously inside the working class by dependence and inside the capitalist relation by function. Supervisors occupy a weaker version of the same duality, and highly credentialed experts another, combining dependence with scarcity rents from their certified skills. The concept matters because it answers the objection that modern organisations are filled with people the simple two-class map cannot place. They can be placed, at the cost of admitting that some locations are contradictory rather than pure, and the resulting map, with poles and theorised positions between them, explains the divided loyalties of the managerial and professional strata better than either a flat continuum or a rigid dichotomy.

Q: Did Marx predict the working class would become the majority?

Yes, in substance. Marx expected capitalist development to proletarianise society: the intermediate strata of peasants, artisans, and small proprietors would be ground down or absorbed, wage labour would become the general condition, and the working class would grow toward an overwhelming majority confronting a diminishing class of owners. History graded the compound prediction unevenly. The proletarianisation element succeeded beyond anything visible in his lifetime, as the wage relation spread from a minority condition in a world of peasants to the default condition of labour on every continent. The simplification element fared worse, since the division of labour inside firms and states generated managers, supervisors, technicians, and credentialed experts faster than concentration eliminated intermediate positions, leaving a class structure more stratified than the two camp picture suggested. The fair summary is that Marx was right about the direction of the relation and wrong about the homogeneity of its result, which is why the working class can be at once historically predominant and internally divided in ways his political expectations did not foresee.

Q: Is this disagreement about facts or about definitions?

Primarily about definitions, and recognising that is the key to the whole debate. The central facts are barely contested: manufacturing employment fell sharply in the older economies, service and office employment grew, incomes and education rose, class identity and organisation weakened, and wage employment expanded worldwide. Disagreement ignites over what those facts are facts about. Counted on a manual or cultural criterion, they record a class in dissolution. Counted on the relational criterion, they record a class transformed in form and enlarged in extent. Because the parties rarely state their criteria, each experiences the other’s conclusion as a denial of plain evidence, when it is in fact the product of a different instrument. This does not make the dispute unreal or merely verbal. The choice between criteria is a genuine theoretical contest about what class is for, what it should explain, and which map predicts the exposures and conflicts that matter. But that contest is won by argument about criteria, not by accumulating facts that both sides already accept.

Q: How does home ownership change a household’s relation to capital, if at all?

Less than the political rhetoric of a property owning democracy suggests, and differently than either side usually claims. The house is a use value and a store of wealth, not a means of production for its occupier: it produces no income, employs no labour, and cannot be realised without surrendering the shelter it provides. Its main relational effects run through the wage itself. A mortgage converts a large share of future labour power sales into a fixed obligation, lengthening the household’s dependence on continuous employment and raising the cost of job loss, strikes, or career change. Accumulated equity does soften the end of the trajectory, making retirement more secure and giving the household a cushion no renting worker has. In the strict terms of class analysis, ownership therefore deepens the wage relation’s grip during the working years while moderating its consequences at their close, a combination that changes the household’s interests and anxieties considerably while leaving its structural position, dependence on the sale of labour power, exactly where it was.

Q: What should a journalist count before announcing that the working class has vanished?

Three things, in order. First, the criterion: the announcement is meaningless until it says whether class is being measured by occupation, income, identity, or relation to the means of production, and most published versions fail this test on their own terms. Second, the employee share: before reporting a death, count how many people in the society in question live by wages and salaries, since on the relational criterion that count is the class, and in the older economies it stands at or near historic highs rather than in decline. Third, the world: a national manufacturing series cannot carry a global conclusion, and any count confined to one older economy will mistake relocation for disappearance. A journalist who runs those three checks will usually find the story has changed shape. What remains reportable, and it is substantial, is the decline of the industrial form, of class organisation, and of class as a public identity, which can be announced accurately under their own names without borrowing the authority of an extinction that the counts do not show.

Q: Why do service jobs strain the factory image without breaking the relational definition?

Because the strain falls on the picture of work rather than on the relation that defines class. The factory image bundles together features that historically travelled together: manual effort, material products, large workplaces, visible collectivity, and a clear line between those who direct and those who are directed. Service work unbundles them. The product may be intangible, the workplace small or domestic, the effort emotional or intellectual, the collectivity invisible. A definition of class built on the bundle therefore fails service work completely, which is why the manual criterion records a disappearance. The relational definition was never built on the bundle. It asks a single question, whether the person must sell labour power to an owner of the means by which the work is done, and that question receives the same answer in the care home and the call centre as on the assembly line. The image breaks because images are of surfaces. The definition holds because it was always about the structure beneath them, and service employment, properly examined, exhibits that structure in full.

Q: How does the global shift of manufacturing change where the working class is counted?

It moves the centre of the count from the nations where the question is usually asked to the nations where industrialisation actually happened across the relevant decades. When production shifted toward East and South Asia, the industrial working class did not evaporate; it was reconstituted, on a far larger population base, among workers drawn from the countryside into factories, construction, and the urban services that grew around them, all selling labour power for a living. A count that stops at the borders of the older economies therefore records a disappearance that is an artefact of the frame. Widen the frame and the same decades show the largest expansion of wage labour in history, with the class’s demographic centre now lying in Asia rather than in the North Atlantic world. The shift also links the two populations materially, since the service worker in the older economy and the assembler in the newer one frequently work inside the same firm’s division of labour, occupying connected positions in a single structure that national statistics describe as two unrelated worlds.

Q: What separates the precariat argument from a claim about job insecurity alone?

The precariat argument claims a class, and a claim about job insecurity alone claims only a condition, and the distance between those is the whole substance of the dispute. Guy Standing’s thesis holds that the precarious constitute a distinct class in the making, with a characteristic relation to production, a distinctive consciousness, and political tendencies unlike those of the old working class, including a susceptibility to movements that promise security through exclusion. For that to hold, insecurity must be more than a bad term inside the wage relation; it must constitute a separate position with interests of its own, opposed to those of secure workers. The critics’ counter is that the evidence shows a gradient of security inside one relation rather than a boundary between two, that casual labour was the historical norm the secure decades interrupted, and that a stratum defined by a condition it shares in degree with millions cannot be expected to act as a class apart.