Baumol's cost disease

A reader's summary of the idea that explains why televisions get cheaper every year while tuition, childcare and hospital stays get dearer: where it came from, how the mechanism works, why a single chart turned it into one of the most argued-over economic ideas online, what it does not explain, and why it has become central to the debate about what AI will do to prices.
The idea at a glance
Baumol's cost disease (also called the Baumol effect) is the tendency of services with little room for productivity growth to become steadily more expensive relative to goods whose production keeps getting more efficient. The stagnant sectors have to pay competitive wages to keep their workers, those wages rise with productivity in the rest of the economy, and so their costs rise without any matching gain in output. It is a cousin of Jevons paradox, which asks what happens when something gets cheaper, and it sits behind much of the frustration captured in the vibecession: the things that got cheaper are mostly things people buy once in a while, and the things that got dearer are the ones they pay for every month.
Origin
The idea comes from the economists William Baumol and William Bowen, who were asked by the Twentieth Century Fund to study why orchestras, theaters and opera companies kept running deficits. Their 1966 book, Performing Arts: The Economic Dilemma, made the point with a famous example: performing a Mozart string quartet takes the same number of musicians and roughly the same time as it did in the eighteenth century, while the number of workers needed to make most manufactured goods had fallen dramatically. A year later, in a 1967 paper in the American Economic Review, “Macroeconomics of Unbalanced Growth: The Anatomy of Urban Crisis”, Baumol generalized the argument to the whole economy, dividing it into a “progressive” sector where productivity grows and a “nonprogressive” one where it does not, and applied it to the rising cost of city services such as schools, policing and sanitation.
History and context
For decades the idea was mostly known to economists. It moved into wider debate as health care and higher education came to absorb a larger share of spending in rich countries, and Baumol returned to it in his 2012 book The Cost Disease: Why Computers Get Cheaper and Health Care Doesn't, arguing that rising health and education costs were largely unavoidable and, importantly, affordable, because the productive sectors would keep making the economy richer. In 2019 Eric Helland and Alex Tabarrok's short book Why Are the Prices So Damn High? argued that the Baumol effect accounts for most of the rise in American education and health costs, more than administrative bloat or regulation.
Others pushed back. Scott Alexander's widely read 2017 essay “Considerations on Cost Disease” asked why, if wages were the driver, teachers and doctors did not feel much richer, and pointed to rising administrative costs, defensive spending and regulation as the parts the theory leaves out. That exchange, between the “it's mostly Baumol” and “it's mostly bloat” camps, became one of the longest-running arguments in online economics writing.
Main ideas
Some work cannot be sped up
A factory can make twice as many shirts per worker over a generation. A string quartet still needs four players and about the same forty minutes. In teaching, nursing, haircuts and live performance the labor is largely the product, so output per worker rises slowly or not at all.
Wages follow the whole economy
Musicians, teachers and nurses could leave for the productive sectors, so their pay has to rise roughly with everyone else's. Their wages climb even though their productivity does not, and the cost per unit of their service rises with them.
Rising prices can be a sign of success
The cost disease is a side effect of growth elsewhere. Services look more expensive because goods became so cheap. A society can afford the dearer services precisely because it is richer overall.
The stagnant share keeps growing
Because the slow sectors get relatively pricier and people still want them, they absorb a rising share of spending and of employment. Baumol predicted this shift toward health, education and personal services decades before it showed up in the statistics.
It is a relative price story
The disease does not say that services get worse or that anyone is being cheated. It says their price rises relative to goods. Whether a given price rise is Baumol, regulation, market power or bloat is an empirical question the theory alone cannot answer.
In the feed
The idea's online life runs through a single picture. Starting in the late 2010s the economist Mark Perry, then at the American Enterprise Institute, regularly published a chart of US consumer price changes by category since the turn of the century, later nicknamed the “chart of the century”. Lines for televisions, software, toys and clothing fall away below zero; lines for hospital services, college tuition, textbooks and childcare climb far above average inflation. The chart is reposted constantly, and almost every reply thread reruns the same argument. One camp sees Baumol: the rising lines are labor-heavy services and the falling lines are traded, automatable goods. Another sees policy: the rising lines are sectors with heavy regulation, subsidies or licensing. A third sees both. The chart works online because it is a picture of something everyone has felt, a phone that costs less each year and a daycare bill that costs more, and because it does not say which explanation is right.
Critique
- Productivity is hard to measure in services. A surgery today is not the same product as one in 1970. If quality gains are undercounted, some of the apparent stagnation is a measurement artifact, and the disease looks worse than it is.
- It can excuse inefficiency. Because the theory says rising costs are natural, it can be used to wave away rising administrative headcount, credential inflation or pricing power that have nothing to do with wages chasing productivity.
- The wage channel does not always show up. If the cost rise were all Baumol, the share going to front-line workers should stay roughly constant. In some sectors and periods it has fallen, which points to other causes.
- “Stagnant” is not permanent. Recorded music, broadcasting and online courses each raised the productivity of performers and teachers. Sectors once assumed to be immune to automation can be reorganized, and the line between the two sectors moves.
- Affordable on average is not affordable for everyone. Baumol's reassurance that society can pay for dearer services holds for the aggregate. A household whose income has not tracked productivity growth still faces the higher price.
Why it matters now: AI
The cost disease has returned to the center of debate because of AI. If models and agents sharply raise productivity in coding, writing, analysis and customer service, the theory predicts that the work they cannot touch, care, hands-on trades, in-person teaching, live experiences, will become relatively more expensive and absorb a larger share of spending and jobs. Several economists have pointed out a sharper corollary: the growth an economy can achieve may end up limited not by its fastest sectors but by its slowest, since the share of spending going to the sectors that do not speed up keeps rising. In that reading, the long-run payoff of AI depends less on how good the models get than on whether the stubborn, labor-heavy parts of the economy can be reorganized at all. It connects directly to the argument in Amara's law about overestimating technology in the short run, and to the question of where the margin ends up when one input becomes almost free.
Impact
Baumol's cost disease changed how economists read price data. It is why statements like “college costs have tripled” now invite the follow-up question “relative to what?”, and why rising service prices in a rich economy are not automatically a sign of waste. For a reader, the useful habit is to treat it as a baseline rather than a verdict: some part of any rising service price is the disease, and the interesting question is how large the remainder is, and what is driving it.