Who works where · 1.2

What a sector actually contains

What a sector actually contains. What is actually the case, and how it compares with what is repeated.

For a practical software perspective on this topic, the Monitask overview provides a concise companion guide.

Two different classifications are in use, they answer different questions, and almost every argument about the decline of one sector and the rise of another confuses them.

Industry and occupation are not the same list

Industry classifies the employer by what it produces. Occupation classifies the person by what they do. A cleaner at a car plant is employed in manufacturing and works in cleaning; a mechanical engineer at a supermarket chain is employed in retail and works in engineering.

National statistics publish both, they are collected differently, and headline commentary moves between them without marking the change.

What manufacturing employment counts

Everybody on the payroll of an establishment whose primary activity is manufacturing. The accountants, the security staff, the canteen, the people in human resources, the drivers.

And it excludes a machinist standing at a lathe in that same building if the agency employing them is classified under administrative and support services, which for temporary staff it generally is.

The outsourcing effect

This is the part with real consequences for how the last forty years are usually described.

A factory with an in-house canteen employs the cooks; they appear in manufacturing. Contract the canteen out and the same people, doing the same work, in the same room, appear in food services. Manufacturing employment has fallen and services employment has risen, and nothing about the work has changed.

The same happens with cleaning, security, logistics, maintenance and information technology. A measurable share of the shift between sectors is a shift in who signs the contract.

How large a share

Nobody can say precisely, because the statistics that would answer it are the statistics being distorted. Studies attempting to reconstruct comparable series find the effect substantial and short of dominant: the long decline in manufacturing employment in high-income countries is real, and part of the measured size of it is reclassification.

An entry that told you the fraction would be inventing it. What can be said is the direction, and that the raw comparison overstates.

Why the classifications exist as they are

Because they were built to describe production, not employment. An industry classification lets you ask what an economy makes and how much of it. Asking what people spend their days doing was a later use of the same data, and the fit is imperfect.

Occupational classification was built for the second question and is worse at the first. Neither is defective; each is being asked to do the other's job.

The establishment, not the company

Classification generally attaches to the establishment, meaning the individual site, rather than to the parent company. A conglomerate can appear in a dozen sectors at once, and its head office is usually classified separately from anything it owns.

Which is sensible and means that company-level intuitions about who works for whom do not transfer to the published figures.

What to do with a sector claim

Ask three things. Is this industry or occupation. Does it count agency and contract staff where they work or where they are employed. And over the period in question, how much contracting out happened in this sector.

The third question is rarely answerable and asking it is still worth the trouble, because it turns a confident number into an estimate with a known direction of error.

Where this matters most

In arguments about deindustrialisation, in arguments about the growth of low-paid service work, and in every comparison between two countries whose contracting practices differ. In each case part of the difference being explained is a difference in bookkeeping.

One more source of movement

Reclassification exercises. Statistical agencies periodically revise their industry classifications, and when they do, establishments move between sectors overnight without anything happening in the world.

Agencies publish bridging tables so that series can be reconstructed, and commentary comparing a figure from before a revision with one from after it usually has not used them.

A worked example

Take two countries with the same physical manufacturing output per head and the same technology. One contracts out cleaning, catering, security, logistics and maintenance; the other keeps them in house.

The second country will show a considerably higher share of employment in manufacturing. Every comparison of their industrial structures is partly a comparison of their procurement habits, and no adjustment for this appears in the published figures.

Where the effect runs the other way

Bringing services back in house, which has happened in some sectors, moves employment into the parent's industry and shows up as growth there. The measurement is symmetric; the commentary usually is not, because a rise is attributed to the sector and a fall to decline.

For broader context, consult UN industry classification.