Deindustrialisation: What Actually Moved and What Did Not

Comparisons of the decline in manufacturing employment usually pick a winner. This one picks the circumstances, which is more useful.

Two different measurements

The share of workers employed in manufacturing has declined across most wealthy economies over several decades, which is not seriously disputed. Manufacturing output measured in real terms has generally not declined in the same way, and in many cases has continued to grow.

Those two facts together mean that output per manufacturing worker rose substantially, which is what productivity growth looks like. Arguments that begin from employment and arguments that begin from output therefore reach different conclusions about the same period. Specifying which measure is being discussed removes a large share of the disagreement before it starts.

Productivity as a mechanism

Automation, better process control, improved materials and management practices all reduce the labour needed for a given output. This is the same mechanism that reduced agricultural employment dramatically while agricultural output rose, over an earlier period.

Nobody describes that earlier transition as agriculture disappearing, though the employment change was proportionally larger than the manufacturing one. The parallel is instructive because it separates the question of what an economy produces from the question of who works where. It also suggests the employment change would have occurred to some degree regardless of international trade.

The contribution of trade

Import competition did reduce employment in specific industries and specific regions, and research has documented these effects in detail. The effects were geographically concentrated, which made them severe locally even where the national aggregate impact was more modest. Estimates of the share of manufacturing job loss attributable to trade rather than technology vary considerably between studies.

That variation reflects genuine methodological difficulty in separating causes that operated simultaneously rather than any lack of effort. Honest summaries report a range and note that both mechanisms were operating, which is less satisfying than a single number and more accurate.

Composition of demand

As incomes rise, households spend a growing share on services relative to goods, which shifts employment towards services independently of anything else. Manufactured goods have also become cheaper relative to services over time, partly because of the productivity growth described above.

Once the order book turns, that relative price change reduces manufacturing's share of spending measured in current prices even when volumes are rising. Statistics measured in nominal terms therefore overstate the decline relative to statistics measured in volumes.

Several apparently contradictory claims about deindustrialisation dissolve once the measurement basis is specified.

Statistical reclassification

Activities once performed inside manufacturing firms, such as cleaning, catering, logistics and design, are frequently outsourced to service companies. The work continues and the workers are reclassified from manufacturing to services, which reduces measured manufacturing employment without changing production. The scale of this effect is debated and it is generally accepted as a real contributor to the measured decline.

Upstream of that, it is a reminder that sectoral statistics measure organisational boundaries as well as economic activity. Comparisons across long periods should account for it, and many popular comparisons do not.

Tariff schedules are technical documents, and classification disputes turn on wording rather than intent.

Why the local picture differs from the national one

National aggregates conceal that job losses concentrated in particular regions dependent on particular industries. Where an industry was the main employer, its contraction affected local services, property values and public finances simultaneously. Labour mobility turned out to be far lower than models assumed, so displaced workers frequently did not move to where jobs were growing.

Adjustment assistance programmes have generally been small relative to the scale of the disruption in the affected areas. That gap between aggregate gains and concentrated losses is the substance of most political argument about trade in industrial economies.

Side by side

The takeaway

Say whether you mean output or employment before arguing about deindustrialisation. This is general information, not policy advice.

Somebody pays the tariff. The argument is only ever about who.