Most of what is said about work is said about a minority of jobs.

Why a share bar on every page

The claim this site is organised around is easy to make and easy to forget, so it is printed at the top of every entry as a measurement rather than a slogan. Before an entry argues anything it says what proportion of employment it concerns, and where that proportion is small it says so and continues.

The figures behind those bars are contested and the definitions behind them disagree with one another. A rough number stated openly is more use than a question left unasked, and the entry on the deskless majority sets out which definitions produce which figure.

What each part is doing

Who works where establishes the composition: what a sector actually contains, how much self-employment there is and what the word covers, how large the gig economy is when measured rather than described, and the work that no survey counts at all.

Time takes the working week apart. Where the eight hours came from, which workers the standard governs and which it has never touched, what the four-day week trials actually measured, and why predictability of hours matters more to most people than the number of them.

Pay treats wage setting as the process it is: a sequence of decisions made by identifiable people under rules, including minimum wages and what four decades of evidence have done to the argument about them, pay transparency, piece rates, tipping, and what each different measurement of the pay gap is measuring.

Hiring asks what each filter selects for, which is rarely what it is advertised as selecting for: the curriculum vitae, the tracking system, the interview and its predictive record, references and what they may lawfully contain, and the error rate of background checks.

Measurement concerns what can be counted, what is counted instead because it is easier, and what happens to any measure once it becomes a target. It includes monitoring and what monitoring establishes, which is less than is claimed for it.

The numbers everyone repeats takes the figures in general circulation one at a time and traces each to the study it came from, including the claim that most jobs in 2030 do not yet exist, the automation percentages, and the engagement figure quoted in almost every presentation about work.

The claim, stated carefully

It is not that desk work does not matter. It is a large minority, a disproportionate share of income sits in it, and changes in it move economies. The claim is narrower and it is about attribution: a statement made about desk work and presented as a statement about work is wrong in a way that is hard to see, because nothing in the sentence marks the substitution.

The correction is a qualifier, not a rebuttal. Say which jobs, and most such statements become true and considerably less exciting. The office is not obsolete; a particular kind of office is less used by a particular kind of worker. Commuting has not ended; the traffic composition changed. Flexibility has not become standard; it became available to people it was already most available to.

There is a test that takes two seconds. Substitute a specific job and see whether the sentence survives. A ward nurse, a delivery driver, a line cook. Most confident claims about the future of work do not survive it, and the ones that do are the ones worth reading twice.

Why a filmmaker

This address carried documentary work and then carried political writing, and now carries this. The films remain listed because they are cited elsewhere and a record should not break; nothing else from either earlier version is republished here.

What carries over is a habit rather than a subject: spending a long time with a small number of questions, and preferring what can be established to what can be asserted quickly. Employment is a subject with a great deal of available evidence and an unusually large gap between that evidence and what circulates about it.

On the machines

Predictions about automation have a long record and it is not a good one. The failures are instructive rather than embarrassing: they mostly come from estimating what a technology can do to a task and treating that as what it will do to a job, when a job is a bundle of tasks held together by coordination, judgement, responsibility and the physical presence of a person.

This site treats each such claim the same way it treats any other figure. What was measured, on which occupations, by whom, and what happened to the previous generation of the same prediction. That is a duller approach than forecasting and it has the advantage of being checkable.

Where the entries come from

Labour force surveys, published research, the text of the legislation where a rule is at issue, and the original reports behind figures that have entered circulation. Where a claim rests on a single study, the entry says so; where the studies disagree, it sets out the disagreement rather than reporting the convenient half of it.

Definitions do most of the work in this subject and most of the confusion. Two honest sources will give different figures for the same quantity because they are counting different things, and an entry that quotes a number without saying what was counted has not told you anything.

The one habit worth taking away

When you next meet a confident sentence about work, ask which jobs it is about. Most sentences survive the question with a qualifier attached, and the qualifier is usually the interesting part.

Who this is for

Anybody who has to make a decision about work and would rather make it against evidence than against a figure somebody read on a slide. That includes people managing others, people negotiating for themselves, and people writing policy who have noticed that the studies rarely say what the summaries say they say.

No prior knowledge of labour economics is assumed. Where an entry uses a technical term it defines it in the sentence it first appears in, and where a dispute among researchers is unresolved the entry says so instead of picking a side and presenting it as settled.

What this site is not

Not career advice, and not commentary on any particular employer. Not legal advice: employment law differs by country and by state and changes often, so an entry that describes a rule names the jurisdiction it applies in and stops there.

It carries nothing over from what was published at this address before. The film pages are the single exception, kept as a record because they are cited elsewhere, and they contain a list of titles and years and nothing else.

For broader context, consult ILO employment resources.