Precarity, Platforms and What Is Actually New

A great deal is written about the transformation of work. Some of it describes real change and some of it describes new words for old arrangements, and telling them apart requires the measurement discipline of Part 7 more than any other lesson in this Topic.

This lesson does that separation. The short version, given first: the average job is not much shorter than it was; the distribution of security has polarised sharply; and the most consequential change is legal rather than technological.

Three workers, one building, one organisation, three employment relationships.

Priya is a nurse, directly employed by the hospital trust. Salaried, pensioned, on a national pay scale, with sick pay, holiday, union representation, redundancy rights and a disciplinary procedure.

Marek cleans the ward. He is employed by a facilities company that holds the cleaning contract. He works alongside Priya every day, in the same uniform-adjacent tunic, on the same wards. He is paid several pounds an hour less than a directly employed cleaner in the same building was paid fifteen years ago, has statutory minimum sick pay, no occupational pension of consequence, and no route into the hospital's grading structure. If the contract changes hands he transfers to a new employer.

And Tomas delivers the hospital's outpatient prescriptions, dispatched through an app. He is classified as self-employed. He has no minimum wage, no sick pay, no holiday, no notice period, and no employer in the legal sense — the app is a platform connecting him to a client.

Three people, one building, one organisation's work, and one of them is legally its employee.

Now the finding that makes this more than an anecdote.

Studies matching workers to firms have found that the same job, performed for a contractor rather than for the client organisation, pays measurably less. For cleaning and security occupations, outsourced workers earn several per cent less than comparable directly employed workers. Analyses of German administrative data following individual workers through outsourcing events — the same person, the same tasks, the same site, a different employer — found wage losses in the region of ten to fifteen per cent, persisting afterwards.

Read that carefully. The work did not change. The person did not change. The building did not change. What changed was which legal entity the wage came from, and the wage fell.

Which tells you what the restructuring is for, and it is not principally about flexibility or focus on core competencies. It is that a large organisation with a pay structure, a union and an internal labour market cannot easily pay a cleaner less than its scale implies — and a contractor can.

Is work more precarious? The answer depends entirely on the measure, and the measures disagree.

Six things get called precarity, and they behave differently.

Temporary contracts. Involuntary part-time work. Solo self-employment. Zero-hours and on-call arrangements. Subjective job insecurity. And job tenure — how long people actually stay.

Some have risen sharply, some have barely moved, and they vary enormously between countries in ways that track employment law rather than technology.

And the one that most surprises people is tenure.

Average job tenure across rich countries has been broadly stable over recent decades, with modest declines for men in some countries offset by increases for women as their attachment to employment strengthened. The image of a world in which nobody stays anywhere is not supported by the tenure data.

Which does not mean nothing changed. It means the change is not where the vocabulary suggests, and locating it requires better measures.

What the data show

Five findings, stated with their variation.

One — temporary contracts vary by country by a factor of five, and the variation is legal.

Some European countries have temporary employment shares approaching a quarter of the workforce; others are nearer one in twenty. The difference is not industrial structure. It is employment protection design.

Where dismissal from a permanent contract is very costly and temporary contracts are cheap to end, employers use temporary contracts for everything — producing a dual labour market in which a protected group of insiders coexists with a large group of outsiders cycling through short contracts, disproportionately young, and disproportionately unable to obtain the credit, housing or stability that a permanent contract confers.

This is a designed outcome of a specific reform in each case, not a technological trend — and the countries that have reduced it did so by changing the law.

Two — zero-hours and on-call arrangements have grown from very low bases in some countries and are essentially absent in others. The measured share is small — a few per cent of employment where it is largest — and its distribution is highly concentrated in specific sectors and among young, older and student workers.

Three — solo self-employment has grown, and a share of it is what is sometimes called dependent or bogus self-employment: a person with one client, working under direction, with no capacity to set prices or subcontract — economically an employee, legally not.

Four — subjective insecurity has risen more than the objective measures have.

This gap is itself a finding, not a measurement error. People report feeling less secure than the tenure data implies they are — and the plausible explanations include the visibility of restructuring, the collapse of the internal labour markets that once made a career legible (see 9.3.2), and the individualisation of risk described in 8.2.3.

Five — and the measure that captures what people actually experience is income volatility rather than employment status.

Analyses of high-frequency bank account data show that a large majority of households experience substantial month-to-month fluctuations in income, with a significant minority experiencing swings of a third or more between months — and this occurs among people in continuous employment.

Variable hours, variable shifts, variable overtime and variable commissions produce insecurity without any change in employment status. A person can hold the same job for a decade and be unable to predict next month's income.

Which is the honest summary of this literature. The average job is roughly as long as it was; the distribution of security has polarised; and the character of insecurity has shifted from the risk of losing a job to the unpredictability of holding one.

The fissured workplace

The change that matters most, and it is legal rather than technological.

Large organisations have shed employment.

Functions once performed by employees are now performed by contractors, franchisees, temporary agencies, supply-chain suppliers and the self-employed. The organisation retains the specification of standards — what must be done, to what quality, at what price, by when — and sheds the employment relationship.

The consequences are structural rather than incidental.

Wages fall, as the evidence above establishes, because the outsourced worker is no longer inside a pay structure that would otherwise apply to them.

Accountability diffuses. The entity setting the standards is not the employer; the employer cannot alter the standards; and when something goes wrong — an injury, a wage violation, a safety failure — the party with the power to prevent it has no legal relationship with the person harmed.

The internal labour market disappears. The cleaner in the story has no route into the hospital's grading structure, because he is not in it. The ladder that 9.3.2 described as bureaucratic control — the mechanism that made the organisation worth being loyal to — has been removed for a large share of the workforce, and it was removed from the bottom.

And competition is relocated. Contractors compete on price for a specification set by the client, and the largest cost they control is labour. The pressure that would once have been a negotiation between an employer and its workers becomes a tendering process in which nobody negotiates with anybody.

This is the single most consequential change in the organisation of work in recent decades, it required no new technology, and it is almost entirely absent from the popular literature about the future of work.

The precariat: the argument and the objections.

The claim is that these developments have produced a distinct class in formation — people characterised not merely by low income but by the absence of a set of securities: stable employment, predictable income, occupational identity, skill reproduction, representation, and a legal status equivalent to full citizenship in the labour market.

Its strengths. It names a real bundle of insecurities that income measures miss, insists that the absence of an occupational identity is a distinct harm, and connects labour market status to political disaffection in a way that is at least suggestive.

And the objections are serious enough that this course must state them.

It is not a class in any of the senses of 8.2.1. Its members share no common relation to the means of production, no common market situation, and no common employment relationship. A graduate on short contracts, a migrant agricultural worker, a redundant industrial worker and a student doing delivery shifts have almost nothing in common except the negative characteristic of lacking security — and a category defined by what its members lack is not a class.

Its members are frequently passing through rather than located in it. For a large share, insecure work is a phase — early career, between jobs, alongside study — which the 9.2.1 developmental cycle point makes visible: a cross-sectional count of insecure workers is not a count of a stable population.

And the framing of the group as politically dangerous — as a mass available for mobilisation by whoever reaches it first — has an uncomfortable lineage, echoing older characterisations of the poor as an unstable and threatening residuum. Whether or not the political claim is right, the framing is the same framing 8.6.2 examined, and it deserves the same scrutiny.

What survives: the bundle of securities is the right thing to measure, and their unequal distribution is real and growing. What does not: that this constitutes a class, or a group with common interests, or a distinct population rather than a shifting one.

Platforms

How large it actually is, and why the argument is larger than the phenomenon.

Estimates of platform work vary wildly across surveys, which is 7.2.2's problem in acute form: the answer depends entirely on whether the question asks about main income, any income, the last year, the last week, and whether renting out property counts.

The consistent finding across the better-designed studies is that platform work as a main source of income is small — commonly one to three per cent of workers in the countries where it has been measured carefully — and that supplementary or occasional platform earning is considerably more common.

So why does it occupy so much of the argument?

Because it makes visible, in a single legible case, a set of arrangements that are widespread and usually invisible. Piece rates, casual hiring, no guaranteed hours, no employment relationship, and payment by output are not new — they describe hiring at the dock gate, homeworking in the garment trade, and agricultural day labour, all of which predate every technology involved.

What platforms did was make those arrangements legible, scalable and photographable, in sectors and countries where they had become unusual. The argument is about employment classification generally, and platforms are the case where it can be seen.

Algorithmic management: what is genuinely new.

Four features, and they are real.

Granularity of measurement. Location, speed, acceptance rate, idle time, route, keystrokes, pauses. Not a supervisor's impression but a continuous record, at a resolution no human observer could produce.

Automation of allocation and evaluation. Who gets which job, at what price, with what priority, is decided by a system — so the decisions that a supervisor would once have made, and been accountable for, are made without a decision-maker.

Information asymmetry. The worker cannot see the rule. They can observe outcomes and infer, and the rule changes without notice. This is the opposite of bureaucratic control (see 9.3.1): a rule that cannot be read cannot be complied with, appealed to, or worked to.

And the absence of anyone to argue with. Deactivation by system, appeal by form, review by another system.

And now the honest correction, because the totalising version is not supported by the evidence.

Workers develop counter-practices, extensively and effectively. Ethnographic and interview studies of platform workers document shared forums, informal knowledge about what the system rewards, coordinated behaviour to manipulate pricing, techniques for managing acceptance rates, and detailed folk theories of the algorithm — some accurate, some not, all functional.

This is 9.3.1's informal system, reconstituted against a system rather than a manual. The control is substantial and it is not total.

And much of it is not new in kind. Control embedded in a system that appears as a technical fact rather than a demand is technical control (see 9.3.2) — the assembly line's logic, with the line replaced by a dispatch algorithm. What is new is the resolution and the reach, not the principle.

And Goodhart's law applies with full force (see 7.2.2). The system measures what is measurable — completion time, acceptance rate, rating — and the workers optimise those, with the predictable consequence that the measured quantity and the desired quality separate.

Classification, which is where the substance is.

The legal question is whether these workers are employees, an intermediate category, or genuinely self-employed — and it determines minimum wage, working time, holiday, sick pay, pension enrolment, dismissal protection, discrimination protection and collective bargaining rights, all at once.

Different jurisdictions have reached different answers, and the trajectory is worth knowing.

A senior court in the United Kingdom held that private hire drivers using a major platform were "workers" — an intermediate statutory category entitling them to minimum wage and holiday pay — on the reasoning that the degree of control exercised over price, terms, route and rating was inconsistent with genuine self-employment, and that the written contractual documents could not be determinative where they did not reflect the reality.

In California, legislation established a presumption of employment, and was substantially reversed for platform work by a subsequent ballot measure funded heavily by the companies concerned.

And the European Union adopted a directive establishing a rebuttable presumption of employment where indicators of control are present, shifting the burden of proof to the platform.

Three things this establishes.

The classification is contestable and is being contested, in courts and legislatures rather than in labour markets.

Written contracts are not decisive where courts look at the substance of control — which matters far beyond platforms, and applies directly to the dependent self-employment described above.

And the outcome is a political result rather than a technical one. The same facts have produced opposite classifications in different jurisdictions within a few years, which means the employment relationship is not a fact about the work but a decision about the work — and 8.8.1's conclusion applies here too.

Because most of what is said about the future of work is a claim about technology, and most of what has actually happened is a change in law and organisational form.

Five things to carry.

Tenure is stable and security has polarised. The average job is roughly as long as it was; the distribution around that average has spread.

The right measure of what people experience is income volatility, not employment status — and it affects people in stable jobs.

Outsourcing lowers wages for the same work in the same building, which is documented by following individuals through the change, and which identifies what the restructuring accomplishes.

Platform work is a small share of employment and a large share of the argument, because it makes an old arrangement legible.

And the classification question is the substantive one. Whether a person is an employee determines a dozen entitlements simultaneously, it is being decided in courts and legislatures, and the same facts have produced opposite answers in different places.

Four questions.

Which measure of precarity is being used? They disagree, and they vary between countries in ways that track law.

Who legally employs the person doing the work? Frequently not the organisation whose work it is.

Can the worker see the rule they are being judged by? If not, this is not bureaucratic control and none of bureaucracy's protections apply.

And is this new, or is it legible? Piece rates, casual hiring and payment by output are very old. The question is why they returned to sectors that had eliminated them — and the answer is in employment law and in the fissuring of the firm, not in the technology.

Three workers in one building doing one organisation's work, of whom one is its employee. And the evidence following individuals through outsourcing events finds wage losses of roughly ten to fifteen per cent for the same person doing the same work at the same site for a different legal employer — which identifies what the restructuring is for.

Six measures of precarity disagree. Job tenure has been broadly stable across rich countries, with modest male declines offset by female increases. Temporary contract shares vary by a factor of five between countries and track employment protection design, producing dual labour markets of protected insiders and cycling outsiders. Zero-hours arrangements have grown from low bases and are concentrated. Solo and dependent self-employment has grown. Subjective insecurity has risen more than objective measures — which is itself a finding. And income volatility is the measure that captures the experience: a large majority of households see substantial month-to-month fluctuations, including people in continuous employment.

The honest summary: the average job is as long as it was, security has polarised, and insecurity has shifted from losing a job to being unable to predict one.

The fissured workplace is the central change and it is legal, not technological. Organisations retain the specification of standards and shed the employment relationship — wages fall, accountability diffuses because the party setting the standards has no relationship with the worker, the internal labour market is removed from the bottom, and competition becomes a tendering process in which nobody negotiates.

The precariat argument names a real bundle of missing securities and fails as a class concept: no common relation to production or market situation, a population many people pass through rather than occupy, and a "dangerous class" framing with an uncomfortable lineage.

Platform work is one to three per cent of main employment where carefully measured, and dominates the argument because it makes visible arrangements — piece rates, casual hiring, payment by output — that long predate the technology.

Algorithmic management is genuinely new in granularity, automated allocation, information asymmetry (a rule that cannot be read cannot be worked to) and the absence of anyone to argue with — and it is technical control in a new medium, workers develop extensive counter-practices, and Goodhart's law separates the measured quantity from the desired quality.

And classification is the substance: a senior court held that control over price, terms and rating was inconsistent with self-employment and that contracts are not decisive; a legislative presumption was reversed by ballot elsewhere; and a European directive established a rebuttable presumption. The same facts have produced opposite answers, which makes the employment relationship a decision rather than a fact.

Fissured workplace — the shedding of employment to contractors, franchises, agencies and the self-employed while retaining the specification of standards.

Outsourcing wage penalty — the measured fall in pay for the same work performed for a contractor.

Dual labour market — protected permanent insiders alongside a large temporary outsider group, produced by employment protection design.

Dependent (bogus) self-employment — a single client, working under direction, without the capacity to set prices or subcontract.

Income volatility — month-to-month variation in earnings; the measure that captures experienced insecurity, independent of employment status.

Subjective insecurity gap — the excess of felt insecurity over what objective measures imply.

Precariat — the proposed class defined by absent securities; a real bundle, and not a class.

Algorithmic management — allocation, pricing and evaluation by system, with continuous granular measurement and an unreadable rule.

Information asymmetry in control — the inversion of bureaucratic control, since a rule that cannot be read cannot be complied with or appealed to.

Employment classification — the legal determination of employee status, which sets a dozen entitlements simultaneously and is being decided politically.

One — count the employers in one building. For any large organisation you can observe, identify how many different legal employers have staff on site. Cleaning, catering, security and reception are the usual answers.

Two — check the tenure data. Look up average job tenure in your country over the last thirty years. Compare it with what you expected.

Three — measure volatility instead. Ask someone in stable employment how much their income varies between months. For hourly and shift workers the answer is frequently large.

Four — try to read the rule. For any app-mediated or metric-managed work, ask the worker what determines their allocation and rating. Note whether they know, and how they found out.

Five — find the classification. For one platform operating where you live, establish the legal status of its workers and whether it has been litigated. The answer will be recent and probably still moving.