Cost disease

Why does a haircut cost more in real terms than a century ago when barbers are no faster?

A barber in 1926 cut roughly as many heads in a day as a barber does now. Nothing about the job has been automated away; the hand, the comb and the head are where they always were. Everything else in the shopping basket has become dramatically cheaper to make. So the haircut should have held its price while the world around it fell — and instead it has climbed against that world.

The usual explanations do not survive contact. Inflation cannot be it: inflation moves all prices together, and this is a relative move. Nor greed, unless barbers acquired a century of extra greed that toolmakers somehow did not. The cause must raise a barber's costs without anything happening inside the barber shop at all.

Ask a plain question: what sets the barber's wage? Not what a barber produces per hour, taken alone. What the shop must pay is whatever it takes to stop the barber going and doing something else, because people can move between occupations. The wage is fixed by the alternatives, not by the scissors.

Now let the car factory next door double its output per worker-hour with new machinery. It can now afford to pay more, and competing factories bidding for the same hands mean it largely must. Wages across the whole labour market drift up in step with what the productive parts of the economy can pay.

What must the barber shop do? Match it, more or less, or watch its barbers leave. And the shop's output per hour has not moved. So its cost per haircut rises by very nearly the whole of the wage rise, and the price of a haircut must climb against the price of a car — driven entirely by an event that happened in the factory.

That is the mechanism, and its force comes from how little it needs. No monopoly, no regulation, no bad behaviour, no change in the service at all. Only two ingredients: uneven productivity growth between sectors, and a labour market that ties their wages together.

William Baumol and William Bowen set it out in 1966 while trying to work out why performing-arts organisations were perpetually in deficit. Their example is unimprovable: a Schubert string quartet written in the 1820s requires four players and about forty minutes today, exactly as it did then. Any productivity gain would have to consist of playing it faster, which is the one thing that would destroy the product.

There is a second consequence, which is why it is called a disease rather than a curiosity. If people go on wanting haircuts and concerts and nursing despite the rising relative price — and by and large they do — then over time a growing share of the workforce ends up in the sectors where productivity is not rising. The economy's measured productivity growth is dragged down by its own success, as the stagnant sector swells.

That last point is the one most worth pulling out, because the phrase "cost disease" invites a gloom that does not follow. A rising relative price is not a rising burden. If output per hour across the economy has multiplied several times over, we can buy more haircuts and more cars than our grandparents could, even though a haircut now trades for more cars than it used to. Baumol spent a good part of his 2012 book insisting on this: the services are becoming dearer against goods and cheaper against nothing in particular, and an economy that has grown can afford them. Most alarm about health care and education "becoming unaffordable" quietly swaps a relative price for an absolute one.

Three honest caveats.

The barber is probably not quite as slow as the story assumes. Electric clippers, booking software and chain-shop scheduling all shave minutes. Service productivity growth is likely positive but small — and genuinely hard to measure, because when the output is a haircut or an hour of teaching, quality changes never show up in the count.

Cost disease is a floor under service-price growth, not the whole of it. How much of the rise in health care and university costs it explains, as against administration, third-party payment, credential inflation and regulation, is actively disputed rather than settled.

And the mechanism depends on the wage link holding. Where a sector's labour supply is separately fed — by migration, by a licensing regime, by work that can be sent abroad — its wages track the rest of the economy less closely, and its relative price rises by less. Whether the premise of the original question is even true depends on the yardstick: measured against a basket dominated by manufactured goods, the haircut has plainly grown dearer; measured against average earnings, far less so.

• Why the same logic makes live performance, nursing, teaching and repair move together in price.
• What it implies for public budgets, most of which buy exactly the services whose relative price keeps rising.
• Whether automation that touches judgement rather than muscle finally breaks the pattern, or merely relocates it.

Every term the collection defines is gathered in the glossary.