Economics
One Hour Is Still One Hour
Machines keep making things cheaper. They can't make an hour shorter. Why care, teaching, live music and your plumber keep getting more expensive, explained with a small town you can run.
Over the past half-century, televisions, phones and clothes have become cheaper next to almost everything else, while a year of college, a day of child care, a night at the hospital and a seat at the symphony have become dearer. The usual suspects are greed, regulation and waste, and each plays some part. But there is a simpler force underneath, first worked out by the economist William Baumol in the 1960s, and some economists argue it explains most of the pattern. You can watch it happen in a town of twelve people.
A town with two jobs
This town has two places to work. At the hat works, a hatter makes two hats an hour. At the barbershop, a barber gives two haircuts an hour, thirty minutes apiece.
Anyone in town can do either job, so the two jobs have to pay the same: two hats an hour. A barber who earns two hats an hour and gives two haircuts an hour must charge one hat per haircut. That is the whole economy.
Now somebody improves the hat press.
Press it a few times and watch the order of events. The press speeds up first, and the hat works can pay more straight away. The barbershop's pay doesn't move, and the barbers notice. Some of them walk. Short of staff, the shop eventually raises its wage to get them back, and it has to raise its price to cover the wage.
After three inventions a haircut costs eight hats. The barber did not slow down. The scissors are the same scissors, and a haircut still takes thirty minutes. Baumol and his colleague William Bowen first made this argument in 1965, about string quartets rather than haircuts.
Nothing happened to haircuts. Everything happened to their alternative.
Count it in minutes
Priced in hats, it looks as though haircuts got dearer. Priced in minutes of human work, you can see what actually moved.
Every way of putting it comes down to one line: the price of a haircut, measured in hats, equals the hatter's output per hour divided by the barber's. Whether you experience that as dearer haircuts, cheaper hats or higher wages depends on what your money happens to track. The ratio is the same in all three.
Same engine, higher stakes
Hats and haircuts carry no baggage, which makes them a good warm-up. Now swap the buildings. The hat press becomes an automated plant. The barbershop becomes a care house, where an hour of attentive care still takes one caregiver one hour.
The chart did not move. Nothing in the mechanism belongs to barbers. It applies wherever a person's time is a large part of what is being sold: teaching, nursing, minding toddlers, keeping watch, playing music to a room, fixing a leak, giving legal advice, checking someone's work.
The bridge
The step most explanations skip is the walk across the street. Care gets more expensive because caregivers could be doing something else, and the something else now pays more. Close the road and see what changes.
With the road closed, care stays cheap, and it stays cheap because caregivers stay poor. The same happens if the machine's gains go to its owners instead of into wages: when the factory wage doesn't rise, the care house has nothing to match. Cost disease is what rising wages look like from inside a job that machines can't speed up.
Richer, and it feels poorer
A care-hour that costs eight gadgets sounds like bad news. Look at what an hour of work buys.
An hour of work still buys an hour of care, exactly as it did at the start, or a pile of gadgets that keeps growing. Care takes a larger slice of the budget because everything around it got cheap. The care bill looks alarming, and the town is materially far richer than it has ever been.
Something has to give
Being able to afford something on average is a long way from agreeing to pay for it. When care gets relatively dear, a town has a handful of ways to respond, and each one moves the strain somewhere else. Pick one and let twelve years run while the machine improves eightfold.
A wage cap doesn't make care cheaper to provide. It turns the cost into vacancies and a waiting list. Holding care spending where it was looks prudent on the spreadsheet and buys a fraction of the care. Paying the full price works for most of the town but leaves the households that gained least priced out; public provision reaches them, and shows up as a large tax bill. Bigger caseloads and robots really do lower the price, partly by changing what is being bought. Only better tools for caregivers lower it while the product stays the same.
Two claims that travel together
So far there has been one claim: the service's relative price rises. That follows from production and wages alone. A second claim often rides along with it: that care, schooling and the arts must end up absorbing most of the workforce. That one needs an assumption about what people want.
If people won't accept an alternative, spending and jobs pour into the service. If a cheap substitute will do (a recording for a concert, an app for a tutor), the human version shrinks. In every case the price chart on the left stays exactly where it was.
Economists have tested both halves. Looking at U.S. industries from 1948 to 2001, William Nordhaus found that sectors with slow productivity growth did see their relative prices rise, as the first claim predicts. The second depends on substitution, as the figure does. In a model by Rachel Ngai and Christopher Pissarides, employment drifts toward the slowest-growing sector when goods are poor substitutes for one another. Georg Duernecker, Berthold Herrendorf and Ákos Valentinyi estimate that American services are good enough substitutes for one another that the future drag from cost disease should be milder than the past one.
Who can afford it?
"Society can afford it" is a claim about averages. If the machine's gains are shared evenly, every household's income keeps pace with the caregiver's wage and everyone can still buy care. If the gains pile up at the top, the average looks fine while the bottom quarter falls behind, a distinction recent work on Baumol's disease has started to take seriously.
Same disease, eight industries
People bring different intuitions to different industries. Nobody is surprised that a concert costs more than a download; plenty of people are surprised by their plumber's bill, or their parents' care home. So here is the same engine run for forty years in eight industries, with the machine improving at the same pace in each.
Where nothing intervenes, the price climbs about tenfold in every one. It climbs less where something gives: classes grow, drafting software takes over part of the work, or the kitchen finally gets faster. The cases also differ in who pays and who goes without, and in whether people will accept a substitute. Each case comes with what studies of the real industry have found.
The lab
Here is the whole engine with every control exposed, and twelve settings to run it in. Each one notes what it teaches and where the analogy strains. Anything you set is saved in the address bar, so you can send a particular town to someone.