Is Baumol's Cost Disease Responsible for the Decline of Public Services?

Across the developed world, people are complaining about the decline of public services. People have to wait longer to get a medical appointment, mail delivery has become more irregular, roads are in worse shape, it takes more time to obtain documentation from the administration, etc. The exact nature of the complaints varies between countries, which makes sense because the range of services that are provided by the government is different in different countries, but there is a lot of overlap and the complaints have the same general flavor everywhere. This suggests that at least some of the causes are the same. A common explanation for that phenomenon is that it’s the result of Baumol’s cost disease. Baumol published a good book on the topic a few years ago, but if you don’t have time to read a whole book about it, the model from which he derived the effect is described in this paper and almost everything you need to know about Baumol’s cost disease is contained in the first 8 pages of that paper. I also really like this paper that Alex Tabarrok recently published on the topic.

The basic idea is very simple yet has proven to have extraordinary explanatory power. Baumol starts from the observation that productivity grows at different rates in different sectors. In order to create his model, he distinguishes between what he calls a “progressive” sector in which productivity increases and a “non-progressive” sector where productivity remains the same, but of course that’s just a simplification. Since wages in the progressive sector will grow with productivity, firms in the non-progressive sector will also have to increase wages or nobody will agree to work for them. But unlike in the progressive sector, in the non-progressive sector, this increase in wages will not be matched by a rise in output since by definition productivity remains constant in the non-progressive sector.¹ Hence the prices of goods and services produced in the non-progressive sector will increase relative to the prices of goods and services in the progressive sector.

Most services tend to fall in the non-progressive sector, while manufacturing is part of the productive sector. That’s because services tend to be labor-intensive and, as Baumol noted, there is often a sense in which the labor that enters into the production of a service is essential to it and even constitutive of the service itself, hence can’t easily be reduced. The most famous example Baumol gives, which is actually what led him to discover the effect in the first place, is live musical performance. While the amount of labor it takes to produce a car has diminished radically over the past century, it still takes as much labor to perform String Quartet No. 14 today as it did a century ago and there is no obvious way to reduce it. Musicians could play it at twice the speed or some could be replaced by a machine, but people would not be interested in such a performance. The same thing tends to be true of other services, such as teaching, which explains why their price has been steadily increasing relative to prices in the manufacturing sector.² In some cases, as the relative prices of some kinds of services have increased, people just stopped consuming them. For instance, it used to be pretty common for people to have servants even if they weren’t super-rich and that’s still very common in poor countries, but now in developed countries only the richest people do.

You can probably see where this is going at this point. The kind of services provided by the government also have low productivity growth, so even if waste and mismanagement in the public sector doesn’t increase, we should expect their relative cost to rise. In general, because of inflation, $100 today will get you less goods and services than $100 in 1950. But the prices of different goods and services don’t rise at the same rate and, in particular, Baumol’s cost disease implies that the prices of services in general and government-provided services in particular will rise faster than prices in general. In other words, even adjusted for overall inflation, the same dollar will get you less public services today than it did in 1950. According to the BEA, the prices of goods and services produced in the US in general have increased at a rate of ~3.2% annually between 1950 and 2023, but the price of goods and services produced or purchased by the government specifically have increased at a rate of ~3.9% during that period.³ This may not seem much, but it means that, even after we adjust for overall inflation, the government has to spend ~60% more to provide the same volume of public services. In other words, because of the change in relative prices, the same inflation-adjusted dollar of government expenditure only buys ~62% as much as it did in 1950.

Of course, we can’t assume that Baumol’s cost disease is the only reason costs have increased faster in government than in the rest of the economy, but it probably explains most of it. In any case, this means that, to maintain the volume of public services per capita at the same level, it wouldn’t have been enough to keep public spending at the same level adjusted for overall inflation. Government spending per capita would have had to increase faster than overall inflation or the volume of public services per capita would have gone down. If that didn’t in fact happen, then Baumol’s cost disease would be a likely culprit for the decline of public services that people complain about. That’s what Frank Lysy, a former World Bank economist, claims in a blog post, but his argument doesn’t make sense to me. He reaches that conclusion by showing that, if you calculate what share of GDP the government would have had to spend in 1950 for the same volume of goods and services if the relative prices had been the same as today, you obtain a figure of ~27%. Since that’s much higher than the share of government spending today, Lysy’s argument goes, it explains why public services have deteriorated so much.⁴

But this argument strikes me as a non-sequitur. Indeed, although the prices of government-provided services have increased more than other goods and services, GDP per capita has also increased a lot during that period, so as long as government spending didn’t decrease too much as a share of GDP this could have more than compensated the impact of Baumol’s cost disease. The only thing Lysy’s calculation shows is that, had the relative prices in 1950 been the same as today, government spending in the US would have had to be much higher as a share of GDP to get the same volume of public services as Americans did in 1950. But the reason why relative prices have changed in that way is that productivity has increased a lot in the rest of the economy since 1950 and, as a result, Americans are much richer today. In turn, the fact that Americans are much richer means that, other things being equal, they can get the same amount of public services with public expenditures that are much smaller as a share of GDP. Thus, Lisy’s counterfactual doesn’t show that Baumol’s cost disease is responsible for the deterioration of public services, it’s just another way of looking at Baumol’s cost disease.

Of course, other things are not equal, because due to Baumol’s cost disease the prices of government-provided services have increased a lot relative to other goods and services and in theory this could have more than compensated for the fact that GDP per capita has increased. So the question is whether the share of public spending in GDP has decreased enough or the prices of government-provided services have increased enough relative to the prices of other goods and services to make the volume of public services per capita go down between 1950 and 2023. Since the share of public spending in the US hasn’t really changed during that period, it went from ~16.8% in 1950 to ~17% in 2023, the question is whether Baumol’s cost disease was bad enough to not only nullify the effect of economic growth but even reduce the volume of public services per capita.⁵ The answer is no. It’s true that, because the change in relative prices, even in inflation-adjusted dollars the government has to spend ~60% more today to get the same volume of goods and services than in 1950. However, since GDP per capita increased even more during that period (it was multiplied by more than 4), the per capita volume of goods and services bought or produced by the government has increased by a factor of ~2.6 during that period.⁶

So how come people feel that public services have deteriorated? I can think of several, non-mutually exclusive possible explanations. First, it could be that the deflators produced by the BEA, which I used to compute the change in the relative prices of the goods and services produced and purchased by the government, are incorrectly estimated. Calculating deflators is very complicated, perhaps especially for the goods and services produced and purchased by the government, so in theory it could be that the prices of those goods and services have increased faster relative to prices in general than implied by the deflators calculated by the BEA. In practice, however, the BEA would have had to get things seriously wrong for the deflators it computes to imply that the volume of goods and services bought and produced by the government have increased by a factor of ~2.6 when in fact it has decreased. I guess one could argue that small errors would compound over time and could therefore result in large mistakes over time, but the BEA employs very competent people who are generally very diligent in their work and I don’t think we should reject the deflators they construct without very good reasons, so I don’t think that explanation is plausible.⁷

Another possibility is that changes in the geographic distribution of the population have increased the inequality of access to public services. On that hypothesis, public services have not declined uniformly but only for some people, who live in areas that have become less well provisioned in public services. The impression that public services have deteriorated comes from the fact that it has actually happened in some areas. Specifically, because people are increasingly concentrated in large metropolitan areas, population density has decreased in small towns and rural areas. The problem is that, when population density goes down somewhere, the provision of many public services in that area becomes more costly. That’s because it often has large fixed costs that become harder to sustain with a smaller population. For instance, many hospitals or schools in rural areas are closed because, due to population decrease, they have become underutilized and as a result the costs per patient or pupil have exploded. If they weren’t closed, this cost increase would presumably be reflected in the deflator for government spending, they often are because it’s politically easier to close them than to increase taxes to finance their higher cost. This seems very plausible to me and I think it’s almost certainly part of the story.

Another hypothesis I find plausible is that the composition of public expenditures has changed in ways that, due to Baumol’s cost disease, have resulted in the deterioration of some public services even if the overall volume of goods and services produced and purchased by the government has no decreased. Indeed, while I’ve been talking indiscriminately about “public services” and “goods and services produced and purchased by the government”, public expenditures are actually very heterogeneous and there is no reason to think that all categories of public expenditures are affected by Baumol’s cost disease to the same extent, because there is no reason to assume that productivity has changed at the same rate for all of them. If public spending on the different categories of goods and services has not changed in line with changes in relative prices, but has increased less quickly for goods and services whose prices have increased more rapidly, then the volume of some goods and services produced and purchased by the government could have fallen even if overall the volume of goods and services produced and purchased by the government has increased.⁸ This hypothesis can only be tested by disaggregating and looking at how different components of public spending have changed over time, but due to political factors I will briefly discuss below, I find it plausible and suspect it’s also part of the story.

One last possibility I want to discuss is that, even if the volume of public services per capita has not decreased for any category, the demand for public services has increased even more than the supply because people want more public services as they become richer. As Random Critical Analysis convincingly argued, this is certainly true for health care, which even in the US is largely provisioned by the government. In fact, in the case of health care, it even seems that the demand for it increases faster than income. In other words, according to that hypothesis, even if the volume of public services per capita had not decreased, people would still feel that it has because the gap between how much public services people want and how much they’re getting would have increased. I also find that hypothesis plausible because people do want more goods and services, whether provided by the government or not, as they get richer and there is no reason to think they would be satisfied by a mere stabilization of the volume of public services they get or even a large increase that nevertheless falls short of the rise of their income. In fact, precisely because people don’t pay directly for public services but only indirectly through taxes (which also means that most of them don’t pay the full cost for them because taxation is progressive), I think we should expect the demand for public services to be less sensitive to changes in relative prices in the case of services provided by the government than in the case of services they buy on the market.

It’s not surprising that, although the cost of goods and services produced and purchased by the government has increased faster than costs in general, the overall volume of public services per capita has nevertheless increased, even if that may not be true for some of them, because GDP has increased even faster and the share of public expenditures in GDP has stayed roughly the same. Indeed, the very reason why Baumol’s cost disease exists in the first place, namely productivity increases, also means that we can afford it. In other words, it’s not really a disease, but merely a side effect of the enrichment made possible by modern economic growth. The real danger Baumol’s cost disease poses with respect to public services is not that it will make them unaffordable in principle, but that it will make them unaffordable in practice because, since that phenomenon is somewhat counter-intuitive and people don’t understand it, it will make it harder politically to fund public services. In other words, it’s less an economic problem than a political one, in the sense that Baumol’s cost disease makes the political economy of funding public services more difficult to navigate.

For instance, even if there were no waste and mismanagement and people didn’t want more health care of higher-quality when they become richer (not to mention the fact that aging increases the demand for health care even keeping income constant), Baumol’s cost disease would still require that we spend more, even adjusted for inflation, on health care to get the same volume of it. But since people don’t understand Baumol’s cost disease, which is not surprising because again it’s pretty counter-intuitive and not well-known outside of people who are interested in economics, when they see that government keeps spending more on health care and that in spite of that the quality of services in hospitals doesn’t seem to improve or is even getting worse by some metrics, they assume that waste and mismanagement is responsible. In turn, this makes it politically more difficult to keep increasing public spending on health care at a rate sufficient to keep up with the increase of the relative price of health care, because people don’t like their taxes to be used to finance what they perceive as a consequence of waste and mismanagement.

Again, it’s not that in theory we couldn’t afford dealing with the consequences of Baumol’s cost disease, but the fact that we can afford to deal with them in theory doesn’t mean that we’ll be able to deal with them in practice because the change in relative prices also changes political incentives in ways that make that more difficult. I think Baumol’s cost disease, together with aging (which unlike Baumol’s cost disease directly affects the government’s ability to satisfy the demand for public services by reducing government revenues and increasing outlays as opposed to just changing political incentives), is already putting the budget of the government in a squeeze and that’s probably going to get even worse over time. Of course, there is plenty of waste and mismanagement in government and we should also try to reduce it, but I don’t think the amount of reduction in waste and mismanagement we can realistically achieve will be enough to prevent Baumol and aging from putting the budget in a squeeze. I think what is likely to happen is that the government will stop providing a some public services and people will increasingly rely on the market to obtain them, which may be good in theory if private companies can provide the services in question more effectively, but will increase the inequality of access to them unless that is accompanied by transfers to low-income households.

1. In fact, as Baumol pointed out, one doesn’t have to assume that wages increase with productivity to derive the effect. What is really essential is only that wages increase in the non-progressive sector in response to the rise of wages in the progressive sector. Other assumptions are necessary in a more realistic model, which also includes non-labor cost, but this needn’t concern us here. What matters is that one can still derive the effect even in a more realistic model and only its magnitude will be affected.

2. Of course, this isn’t true of every service and, even for something like teaching or live musical performance, there is some productivity growth. For instance, as Baumol observed, faster and cheaper transportation allows musicians to play the same quartet in more places during the same period for a smaller cost. But it’s still true that productivity growth tends to be much slower in the services sector and that’s all we need to derive Baumol’s effect.

3. I calculated the annual rates based on the deflators I found in table 1.1.9 of the National Income and Product Accounts published by the BEA.

4. Lysy actually looked at the period from 1952 to 2012, because he wrote his blog post in 2013, but one can make the same argument for the period between 1950 and 2023 that I’m looking at in this post. This is what I did and that’s why the figures I give are somewhat different from Lysy’s.

5. I calculated the share of public spending in GDP in 1950 and 2023 from the data I found in table 1.1.5 of the National Income and Product Accounts published by the BEA. For some reason, Lysy says that it has decreased from 25% to 20% between 1952 and 2012, but he doesn’t give a source for that claim and that’s not what the data published by the BEA that he relies on elsewhere in his post show.

6. I did that calculation based on the data on GDP, government expenditures and deflators in the tables 1.1.5 and 1.1.9 of the National Income and Product Accounts of the BEA and the data on population from the US Census Bureau.

7. Strictly speaking, it’s actually misleading to talk as if there were a correct deflator for government spending, because the construction of a deflator necessarily involves subjective choices, especially to account for changes in quality, so it’s not even as if there were a unique right way to do it. But this doesn’t mean that it’s entirely subjective and that some ways to do it aren’t wrong.

8. Another way in which a change in the composition of public expenditures could have resulted in a decline of the provision of public services even if overall the volume of goods and services produced and purchased by the government has increased would be if public expenditures that depend less on population size, such as defense, had increased more rapidly than public expenditures in general. This would have squeezed other public expenditures and could have reduced the volume of public services per capita. However, my impression is that this has not happened, which is why I only mention that hypothesis in a footnote. Defense spending in particular has decreased a lot since the end of the Cold War as a share of GDP. This is true even if we take 1950, before the spike caused by the Korean War and the implementation of NSC 68, as the reference year, as I’ve done above.