Followup On The Baumol Effect: Thanks, O Baumol

Last week I reviewed Alex Tabarrok and Eric Helland’s Why Are The Prices So D*mn High?. On Marginal Revolution, Tabarrok wrote:

SSC does have some lingering doubts and points to certain areas where the data isn’t clear and where we could have been clearer. I think this is inevitable. A lot has happened in the post World War II era. In dealing with very long run trends so much else is going on that answers will never be conclusive. It’s hard to see the signal in the noise. I think of the Baumol effect as something analogous to global warming. The tides come and go but the sea level is slowly rising

I was pretty disappointed by this comment. T&H’s book blames cost disease on rising wages in high-productivity sectors, and consequently in education and medicine. My counter is that wages in high productivity sectors, education, and medicine are not actually rising. This doesn’t seem like an “area where you could have been clearer”. This seems like an existential challenge to your theory! Come on!

Since we’re not getting an iota of help from the authors, we’re going to have to figure this out ourselves. The points below are based on some comments from the original post and some conversations I had with people afterwards.

1. Median wages, including wages in high-productivty sectors like manufacturing, are not rising

I originally used this chart to demonstrate:

Some people protested this was was a misleading portrayal, and that there are structural factors that disguise rising wages. I’ve written about this before in Wage Stagnation: Much More Than You Wanted To Know. The short answer is – no, it’s not about increasing benefits, those only explain about 10% of the wage-productivity difference.

It is partially about how you calculate inflation, which explains around 35% of the problem depending on who you believe. But we’re comparing wages to the cost of education/medicine. As long as you’re using the same deflator for both of them, you’re fine. As far as I know, T&H and everyone who talks about rising education/medical costs has been using the normal consumer deflator. So if you want to argue wages are underestimated, you also have to argue education/medical costs have gone up even more than people think. This doesn’t help at all!

(or see these numbers, which show that nominal college tuition has gone up as a percent of nominal median wage, and so should be immune to inflation shenanigans)

Other people protested against looking at the median wage, arguing that the wages of college graduates are more relevant. After all, teachers, professors, doctors, and nurses are all college grads. If their opportunity cost goes up, that could still drive a Baumol effect. And:

Surely doctors and professors are in that top blue line; I think nurses and teachers are in the lower green one. Plausibly these professions’ opportunity costs have gone up 50 – 100% during this period. This is a start to explaining why education/medicine have gone up 200 – 300% during the same time. On the other hand, the period of fastest wage growth was 1965 – 1975, which as per T&H’s graph (page 2) was the period of slowest cost growth.

2. Wages for doctors and teachers have not risen

Let’s start with teachers. T&H use NCES’ “instructional expenditures” category to show teacher wages have tripled since the 1950s; I cited NCES’ actual teacher salary data to show it’s stayed about the same. What’s going on?

NCES only has good data after 1990. Their data says education has become about 30% more expensive in real terms since that time – T&H’s data (page 2) suggests it has become twice as expensive, but their data on page 5 agrees with NCES (huh?). Here’s how various fields have changed, using two different classification systems:

Pay attention especially to the first one. From 1990 to 2016, employee salaries as a percent of educational expenditures have gone down! Employee benefits have gone up a bit, but not enough: salary + benefits is still a smaller part of the education budget in 2016 than in 1990. How do you look at these data and say “We’ve figured out why education costs are rising and it’s definitely salaries”?

I think I miscalculated my tone when criticizing T&H’s presentation of data in their book. Tabarrok says I complained about areas “where [he] could have been clearer”. But my actual concern was that the presentation of this section misleads the reader. Whenever T&H talk about something other than salary, they emphasize that its share of the pie has not increased, but don’t mention that it increased a lot in absolute terms. Then when talking about salary, they emphasize that it increased a lot in absolute terms, but don’t mention that its share of the pie hasn’t increased. You’re left with the impression that salaries are the culprit for the price increases, when in fact salaries increased least of all the major categories in the data. “The data could have been clearer” is never just a minor gripe! Unclear data means you can prove whatever you want!

How are salary costs per pupil rising (even in proportion to other costs) if salaries are not? My guess is it’s all about decreasing class sizes, which T&H also highlight.

Moving on to doctors, I don’t have any equally clear sources. But I’ll at least try to explain more about the ones I already have.

This paper from Health Economics, Policy, and Law shows on page 15:

They cite their source as “references 48 – 54”, but their reference section isn’t numbered, and is in alphabetical order, which would make it a pretty big coincidence if reference 48 through 54 were all about the same thing. I think a wire got crossed somewhere. But taking it at face value, I eyeball US doctor salaries in 1960 as 130K and in 2005 as about 210K, for an increase of 60% (not the 200% T&H claim). Doctor salaries have been about stable since 1975, even as (according to T&H’s graph on page 2), healthcare costs have about doubled.

The Last Psychiatrist posts this image:

I can’t trace the source beyond him, but read his post, where he notes that “There is the reality that doctor salaries (with notable exceptions) have been fairly static since 1969, even as the cost of living, price of homes, college, etc have gone up. And medical school debt.”

The last source I was able to find was this 1985 paper on doctor pay. It states that “In 1973, the median annual physician income was $45,000…by 1982, the median physician income was $85,000.” According to the inflation calculator, those numbers are $260,000 and $225,000 in current dollars, respectively. Estimates for average doctor salary today range from $209,000 to $299,000. I’m surprised how hard this is to measure, but it doesn’t seem to have doubled or tripled the way T&H claim (or the way it would have to in order to drive Baumol effects).

Also, the Baumol effect only works if the market sets your salary in the first place! Right now the supply of doctors is limited by licensing issues and bottlenecks in the medical education process; that keeps salaries high. Why would the Baumol effect drive that even higher? If doctors’ salaries didn’t increase in keeping with the highest-productivity industries, would medical schools sit empty? Given that the people setting salaries for doctors (hospitals, clinics) are not the people who determine the supply of doctors (bureaucrats, medical school deans), why should supply and salary be related in a way that obeys normal economic laws? I’m not really sure how to model this, but I’m pretty sure it doesn’t end with doctors leaving medicine to play violin concertos instead.

3. The Baumol effect cannot make things genuinely less affordable, but things are genuinely less affordable

I think I screwed up here.

The Baumol effect cannot make things genuinely less affordable for society, because society is more productive and can afford more stuff.

However, it can make things genuinely less affordable for individuals, if those individuals aren’t sharing in the increased productivity of society.

Suppose that in 1960, widgets cost $1, a worker could produce ten widgets an hour (and made $10 in wages), and violin concerts cost $10. Also, you are a farmer and make $10 per hour. You can listen to one violin concert per hour.

In 2010, widgets cost $0.50, a worker can make a hundred widgets an hour (and makes $50 in wages), and violin concerts have risen to $50. But you, still a farmer, still only make $10 per hour. The Baumol effect has driven up the cost of violin concerts for you.

This shouldn’t happen in real life, because if you work in the high-productivity-gain industries, you should make more because of your increased productivity, and if you work in the low-productivity-gain industries, you should make more because of the Baumol effect. But if it did happen, you’d be screwed.

And in fact, as mentioned above, wages have not increased in keeping with productivity, either in the high-productivity-gain industries like manufacturing, or in the low-productivity-gain industries like teaching. So if you work in one of those industries, it’s totally possible for you to be screwed, ie for college etc to become much less affordable for you.

If wages had grown in keeping with productivity, median yearly salary would be something like $100,000. Someone making $100,000 per year shouldn’t have that hard a time affording health insurance and college tuition for their kids; this would be the Baumol effect working in its normal, rising-tide-lifts-all-boats way. Since this hasn’t happened, Baumol-applicable industries have become harder to afford.

Again, this is all theoretical, because wages in high-productivity industries haven’t risen, so it’s hard to see how a Baumol effect could be happening at all. But if it was, it would be a good explanation for cost disease.

So I retract this third objection. I think the first objection mostly still stands, though it is a little weaker if we limit ourselves to college-educated workers compared to all workers. I think the second objection absolutely still stands, and it’s hard for me to see how T&H’s case could survive it.

This entry was posted in Uncategorized and tagged . Bookmark the permalink.

186 Responses to Followup On The Baumol Effect: Thanks, O Baumol

  1. Ttar says:

    One word: Outsourcing.

    You can outsource the production of goods easily. Less than college educated workers traditionally worked in the production of goods. The price of goods has fallen. The wages of greater than or equal to college workers have gone up. You can’t as easily outsource skilled services, especially medicine and education. This might tie into some kind of substitution effect based on productivity. Unskilled services (food and hospitality, etc.) has given the less than college workers an alternative income stream, but there’s plenty of immigration to keep those wages low. The people who owned the companies who profited from outsourcing had college degrees, so the less than college workers didn’t get a capital-based income stream that lets them price compete for big purchases like college and medicine. The educated wealthy then spend enormous amounts on education and housing because they are positional.

    • 10240 says:

      This doesn’t explain why education and healthcare prices have gone up faster than teachers’ and doctors’ salaries.

      • Cory Giles says:

        Potentially it could, in conjuction with the Baumol effect and a few other things.

        First I note that these cost-disease industries generally seem to have two factors in common: (A) they are services or otherwise subject to the Baumol effect, (B) they are price-inelastic to demand (i.e., they are considered “essential”), and (C) they are not outsourceable.

        So the chain of events I propose is:

        1. Productivity in widget-making increases due to technology. We would expect this to lead to wage increases for widget-makers and services alike (via Baumol), but:

        2. Most of the gains in terms of wages/profits from productivity growth are captured by capital and by increased wages in third-world countries producing the widgets, not workers in the West, because of outsourcing.

        3. However, widgets are still cheaper in the West. This means that while disposable income is flat, a greater proportion of consumer income is available for services.

        4. Prices rise for those essential services in proportion to the income available for them, because they are “essential” and price-inelastic. However, wages for those service providers do not rise in tandem because as mentioned in (1) the gains from productivity rises have gone elsewhere.

        The only real role the Baumol effect plays in this model is that it entangles the wages of domestic widget-makers and service providers so that they don’t get much out of sync. However it does help account for the apparent paradox in the price of services rising without wages rising in tandem.

        In this model we would also expect to see wages only rise when productivity in non-outsourceable, or at least non-outsourced, industries increases. I don’t follow this topic that closely so I’m not sure whether what I have just outlined is simply the conventional wisdom on the subject.

    • Dack says:


      Scott says; “wages have not increased in keeping with productivity”. Presumably that is only within our society. Is it possible that increased wages/productivity overseas are baumoling us?

      • 10240 says:

        The Baulmol effect involves the choices of employees in the affected sectors. Increased productivity of foreigners in manufacturing sectors would only create a Baumol effect if emigrating and becoming factory workers was a serious choice for many Americans. However, that’s not happening, because those foreign workers are still poor compared to American employees (even low-skilled ones, let alone high-skilled), they are only getting richer compared to when they were dirt poor.

    • Charles Engelke says:

      The decline in absolute share of total income for the median worker could be mostly lower market demand for the type of work median workers do. This could be because of technological changes in production, but more probably growth in outsourcing during the period in question was indeed the biggest change.

      However, it is worth pointing out that use of CPI to adjust wages in step with cost of living would also have this effect. Raises linked to CPI change guarantee the continued affordability of the items in the CPI basket. However these items shrink as fraction of the total GDP as consumer goods become cheaper in work exchange terms. So share of total income will shrink automatically. Housing, schooling &services will become harder to pay for as observed here if this is the case.

      Inflation from 1979 to 2015 according to CPI = 3.7x
      Inflation from 1979 to 2015 in terms of constant work/hr = 6

      So a purely CPI adjusted salary is even worse than what happened on average from 1979 to 2015 to the median sector I examined in the link above (reduction to 75% of 1979 work in the relative average pay/hr value):
      The salary in 2015 would be equivalent to only 60% of the original base salary from 1979 supposedly being adjusted yearly to ‘keep up’ with inflation

      If you want all wages to rise with ‘productivity’ equally you must adjust using nominal change in average pay per hour, not CPI!

  2. Alchemist says:

    Thanks for creating this discussion Scott. -off topic- There are some facts of life that are hard for me to understand. The most significant is probably:

    The maths are hard.

    Once a discussion includes inflation, exponential rates, and doubling times and such you probably lose 90% of the general population, and 60-70 percent of SSC’s audience.
    -return to topic-
    Baumol’s effect is what it is. It has no bearing on rising prices of consumer goods. You will have to look elsewhere.

    • BlindKungFuMaster says:

      I think your maths are off. If you lose 90% of the general population you shouldn’t be losing anything close to 60-70% of the SCC readers.

  3. ajfirecracker says:

    Inflation is not a neutral force that affects all money equally and which you can just divide out. Here is an intuitive explanation:

    • RicardoCruz says:

      Was that blog post written by an economist?
      The author of the blog argues that, in the short term, those who receive the new money supply benefit relative to those who only get the new money when it trickles down to them.

      First, it assumes that price inflation can only be caused by money inflation.
      Second, it assumes that people are stupid. In the real world, landowners, businessmen, workers already know more or less what inflation to expect, and already built-in inflation in their contracts. In my country, all rents/wages/prices are adjusted yearly based on inflation, which is very low anyway.
      Thirdly, even if what the blog says is all true – it does not apply to long time spans as are being discussed here, and there is no reason to believe doctors receive the new money supply first in any case.

      PS: Don’t take me wrong. There might be problems in the numbers being discussed because of the way that inflation is calculated. There are many debates to be had about that. But that blog post is weird, sorry.

      • ajfirecracker says:

        Why does it matter if the author is a professional economist? His logic either holds up or it doesn’t, regardless of his credentials.

        The phenomenon he is describing is called “Cantillon Effects” by professional economists if you want to learn more.

        I think you misunderstood my point. I don’t think that doctors get new money first, I think the general public gets new money last.

        • drunkfish says:

          Why does it matter if the author is a professional economist? His logic either holds up or it doesn’t, regardless of his credentials.

          That seems like a perfect example for

          “Blogger writes post that seems compelling but it’s outside my expertise and I’m not sure about it” should have very different impacts on one’s priors depending on who the author is. We aren’t perfect at determining whether arguments are right, so outside of a fairly small subset of arguments, we need to treat compelling but unfamiliar arguments in a Bayesian way rather than “logic either holds up or it doesn’t.”

          • ajfirecracker says:

            The problem with that is that I can provide outright contradictions from legendary and often Nobel-winning economists. Moreover, the logic here is not that difficult for a layperson to grasp (I happen to have academic training in economics, but virtually all issues of real importance to laypeople are addressed in introductory courses, you don’t have to take years of training to understand why e.g. rent control is likely to cause housing shortages and reduction in quality of housing)

        • Matthias says:

          And if you read about the Cantillon effect, you quickly figure out that it’s basically non-existant for expected money creation, exactly for the reasons drunkfish figured out: anticipation.

          At the time the coming money creation is known, prices adjust. Especially in liquid capital markets.

          (So in most countries, central banks provide a guaranteed base line demand for government debt. But the only party benefiting from the fully anticipated increased price of government debt is the government itself who can make new T-bills from scratch.)

  4. Steve Sailer says:

    Basically, over my 60 years of life, manufactured goods have gotten cheaper (in part due to vastly improved logistics), but the basics of bourgeois family life — e.g., higher education for a white collar career, a house with a yard in a decent public school district, a wife who stays home with the children when they are young, and reliable medical insurance — have gotten much more expensive.

    • RicardoCruz says:

      Very true. And that probably explains the reduced fertility.

      • Aapje says:


        This disaggregated inflation graph suggests that children have become much more expensive than aggregated inflation would suggest.

        • SaiNushi says:

          This is in part because safety standards for children’s stuff has made things much more expensive.

          I recently learned that unless you’re willing to shell out $350 for a car seat, you have to get a new one every single time you get in a fender-bender to ensure your child’s safety. And children are supposed to stay in their car seats for much longer than they used to when I was a kid. I currently live in KS, and by their laws I would’ve been in a booster until I was 12 or 13 o.o

          Plus, safety standards for cribs, and strollers, and jumpers…

          • Steve Sailer says:

            E.g., if you have 3 kids, you are often virtually legally obligated these days by child seat safety rules to buy a large 3-row SUV or minivan.

        • Aapje says:

          Oops, forgot the link.

    • ChrisA says:

      Steve, in point of fact home ownership rates were much lower in the 1950’s than today. And houses are much nicer and bigger than then as well.

      I am not quite as old as you but l remember that time very differently. We ate boring crap food, cars broke down a lot, houses were small, dark and cold and most people rented, and my dad was frequently out of work, people died earlier and so on. Life is much much better now for most people even if wives work (sarcasm). Of course there are going to be some downward mobility or people who saw their circumstances decline, probably for you 1950’s California was pretty good, but you can’t generalize this to all society.

      • RalMirrorAd says:

        I wasn’t alive at the time, but my guess is that the increase is a combination of higher expectations for what consumption is reasonable for someone aspiring to be middle class, plus higher expectations for what consumption is mandatory for a parent.

        So if someone wants to be perceived of as being middle class they have to consume levels of housing, healthcare, education, and childcare, that are vastly greater then what existed in 1950, and the improvements in productivity are insufficient [or nonexistent depending on the industry] to compensate.

        There are all sorts of legal/cultural barriers to people opting for smaller homes, 5k/child-year schools instead of 10k-15k schools. Social aspirations drive more people into the workforce, creating the phenomenon of daycare costs, which further pushes people away from fertility and into careerism etc. It feeds on itself.

        I suspect in the 50s and 60s people back then could have attained a relatively higher standard of living by putting off childbearing, but social/cultural norms would have frowned upon that kind of lifestyle. Nowadays the opposite is true, forgoing educational attainment and income and consumption for fertility is seen as a low-class lifestyle.

        This is also why, I suspect, poorer immigrants who aspire to higher standards of living start off with high fertility but second/third generation much less so.

  5. Alex Zavoluk says:

    I think I commented this point in the original thread, but perhaps clarifying it with the authors would help: In my understanding of the Baumol effect, the wages of low productivity wages shouldn’t be increasing very quickly. The Baumol effect describes a situation in which the wages of, say, violinists once were relatively high, but now they are not. That’s the whole point of stipulating that they have low productivity growth to begin with; we expect that their wages don’t go up, and they only really go up because they’re dragged, kicking and screaming, by more productive industries. In other words, flat wage growth for education and medicine sounds like exactly what you would expect, in spite of technology that might improve it.

    • benwave says:

      If this is true, it still doesn’t explain the increased cost of education and healthcare, does it?

      • Clutzy says:

        No, which is why everyone focusing on it misses the point. It only explains why the cost hasn’t dropped precipitously.

        • Alex Zavoluk says:

          If the cost of most things drop precipitously, but the cost of healthcare and education stays flat, then wasn’t one of the main points of the book that this constitutes a price increase in HC and Edu, since prices are relative?

          • Clutzy says:

            That is a second order effect not Baumol, as far as I understand it.

            Baumol is this:

            12th grade teachers in 1960: I educate 20 kids per hour and give them a teaching effectiveness of X. My salary is $10,000

            12th grade teachers in 2010: I educate 20 kids per hour and give them a teaching effectiveness of X. My salary is $40,000

            Wait, what? Why are we paying you 4x as much for the same work?

            But then Baumol explanation is:

            Farmer 1960: I farm 100 Acres of land, each acre produces 10 Units of Corn. My Salary is $10,000

            Farmer 2010: I Farm 400 Acres of land, each acre produces 11 Units of corn. My Salary is $41,000

            In order for there to be teachers (who are as qualified as farmers in this example) in 2010, they had to have almost the same salary increases.

            Anything beyond that is a different effect. The fact that Corn is now $1 instead of $4, means there are extra dollars chasing down things that are not corn. But that is simple supply/demand curves, not Baumol.

          • deciusbrutus says:

            Let’s take those numbers more literally:

            Four teachers in 1960 educate 80 kids and are paid $40,000
            Four farmers in 1960 produce 40 corn and are paid $40,000

            Then a farm improvement occurs.

            One farmer in 2010 produces 44 corn and is paid $41,000
            Four teachers in 2010 educate 80 kids and are paid $160,000?
            Three former farmers become concert violinists and are paid $120,000?

      • Alex Zavoluk says:

        Why wouldn’t it? Baumol is all about relative increases, right?

        • benwave says:

          It’s the difference between cost of service and cost of labour though. The claim of Tarrabok and Helland, if I understood Scott correctly, is that the Baumol effect causes wages to rise in health and education. This, in turn causes the cost of health and education to increase.

          Wages for professionals have risen by perhaps 100% in the time period under discussion. Health and Education costs have risen far more than that during the same time period and using the same deflator (from memory of the original cost disease post, wasn’t it something like 300%?). The wage increases cannot explain the cost increases of purchasing healthcare or education – at least not more than at most about a third of it.

      • benf says:

        It explains part of it, but not all of it. Here is where comparing other countries becomes useful. It helps you disaggregate the general “industrial revolution” effects from other things like rent-seeking and privatization and union busting.

    • David Condon says:

      Yeah, Scott Alexander’s argument is off base here. Wages haven’t gone up… AFTER adjusting for price inflation. But price inflation is primarily affected by increases in wages. So he’s adjusting for the variable that he’s supposed to be investigating.

      • 10240 says:

        The point is that prices in some sectors have increased faster than the general price inflation (CPI) we adjust for.

  6. Flavoryeagle says:

    I hate to basically re post … but I work with a lot of statistics and when I run into a problem that seems to add up in weird fashions I go through specific data points. I find a point that is either convenient or should be representing a random draw.

    It would be a lot easier to work through whether teachers are getting paid the same by looking at equivalent teachers at different times in the same school. The deflated wages in your sample could represent some change in the cohort.

    A teacher might be a broader term today in your stats than previously. There are 100 ways it could end up being off and we will never solve it in our free time.

    (Complete aside, females proportions in the workforce changed drastically over this time period and it would be surprising if wages weren’t suppressed until at least the 70’s due to discrimination against females)

    Now the only single time series I can construct is for college professors in ivy league schools:

    I found a Dartmouth 1998 factbook with comparable college professor wages by rank. ( on pages 81-83) Using the factbook in addition to ( I estimated that these colleges had about 3.75 – 4.25 CAGR on professorial wages from 1989 -2017. ie Professor’s nominal salaries tripled arithmetically in that time period. Full professors at Dartmouth were paid 61k in 1989 and 188k in 2017.

    You can also look at tuition for 1989 and it was 13k and in 2017 it was 51k. That outpaces professor’s wages (which would imply 40k tuition) but not by a lot. I am guessing there are many reasons that is not a 1-1 correlation in the additional tuition but I would honestly believe a large portion is the fact that those are sticker prices and not the actual costs each student is paying. College’s have perfect price discrimination and “tuition” is just .. well you know all that.

    Let’s assume professors’ wages are a good proxy for the population of similarly educated, high ability workers. We can definitely say they grew in excess of inflation in that period by about 1.5% annualized.

    Now we can work back from this time series. Why is this Dartmouth sample so different from the explanation for other college professors? If your stats indicated professor’s salaries grew in line with inflation and it appears they did not, why are those stats trusted for teachers and doctors?

    • JohnBuridan says:

      Dartmouth currently averages ~25k for a Freshman, which is if we assume that 13k was a fairly normal real cost at the time, is exactly the same amount of money. We should assume though that the distribution has changed. Some students pay the extravagant 51k in 2019, which helps float the students who pay 25k.
      25k in 1989 BTW is 51k. What this tells me is that while costs have gone up, institutions have been trying lots of different structures to hide those rising costs, then as more and more compartments of the institution make moves to hide their rising costs it becomes impossible to figure out exactly how much blood work for pregnant women should “actually cost.” And for us, it becomes really difficult to see what the true causes of rising costs actually are since each compartment of the industry hides its costs, makes itself as necessary as possible, and raises its profile.

    • CandidoRondon says:

      I actually went to Dartmouth and fortunately for the college we had a blogger, Joe Asch, who would investigate these kinds of things until he passed away last year.

      The rise in Dartmouth’s tuition can be almost wholly explained by the growth in non-essential administration. For example, Dartmouth, a college of 4000 undergraduates & approximately 2000 grad students had in 1999 about 2400 non teaching staff. By 2015 the college had 3500 non teaching staff. The number of non teaching staff increased ~500 people between 2010 – 2014 while the number of students only increased by 150 and number of professors only increased by 35. At the those rates of growth I imagine by next year there may be a non teaching staffer for each undergraduate. When you realize that each undergraduate is paying for the salary if a well compensated staffer with six weeks of pension and one of the fanciest health care plans in America it’s no wonder tuition has increased as much as it has.

      In addition full professors are only part of the picture – while they continue to maintain their salaries they college now relies on a much greater percentage of assistant professors compared with 1998. The excess compensation is going to 800k a year “Deans” while junior professors have very low salaries (like 80k a year at a place where an entry level busboy can make 40k base + overtime)

      • A Definite Beta Guy says:

        The excess compensation is going to 800k a year “Deans” while junior professors have very low salaries (like 80k a year at a place where an entry level busboy can make 40k base + overtime)

        I understand that this is a nit-pick, but this pretty much resembles a Baumol effect. If the price of labor at Dartmouth has doubled in real terms because the economy is so good there that even Busboys can make $40k an hour, and it drives up prices because 1 teacher>>>>1 class, that’s Baumol.

        It does sound like the labor mix has changed, which is driving the total cost, but if it were just junior professors…well, that’d be your answer.

        • CandidoRondon says:

          Dartmouth pays much higher salaries than any other employer in the region though – that’s mostly due to choice and not necessity since Dartmouth could pay market rates if they wanted to.

  7. Tenacious D says:

    During at least part of this period consumer spending power has risen more than wages would indicate. Interest rates have fallen substantially since the early 80s and household debt has increased at the same time; average debt service payments have only fluctuated between ~ 10 – 13% of household income.

  8. JohnBuridan says:

    I wonder if there is any cost disease in Law or Seminary Tuition. Maybe if we can find edge industries where there is only a little cost disease we can solve the riddle of how little versions of cost disease here and there have led to way larger cost disease overall.

    So we want to figure out how lots of causes work together to create a nonlinear effect. The reason the effect is nonlinear is that any two causes on their own do not create the massive cost disease, but when taken together they do. It’s very hard to isolate all the contributors but here are a few standard causes of market failures:

    Information asymmetry: it’s extremely hard for purchasers and funders of education and health to assess its value.
    Time horizons: the aggregate value of an educated populace can hardly be assessed, and the cost of an uneducated populace is likewise hard to assess until years have elapsed. Similarly, health is over a lifetime, and we don’t know how expensive an intervention will turn out to be until many years elapse.
    Genetic drift: Inability to coordinate for better versions of the product because of the long cycles, and so the medical institutions can float further away from acting as though they are in competition with one another.
    Local monopolies.
    Entrenched Equilibria.
    Pressure to provide more goods, but providing beyond a certain point drives up cost exponentially.

    These are some guesses. But I don’t understand the proper method to sort through them. Any guidance?

  9. sarth says:

    I really have trouble following any of this because of the obvious absurdity that increased productivity of a factory would drive factory workers’ wages up.

    A Marxist might say that it should work that way, but as far as I know no one alleges it does work that way.

    It would only work that way if the increased productivity of the factory worker required more rare skills. I’m reluctant to think that’s been the rule.

    Unless I’m completely missing the point, this makes that whole aspect of the premise nonsense which makes the whole discussion feel like an imaginary story.

    • Alex Zavoluk says:

      That’s not what cost disease/the Baumol effect predicts. Factory workers wages have gone up. The effect predicts that if Doctors do not become more productive, their wages will go up anyway when factory worker wages go up.

    • blumenko says:

      As long as productivity increases across a whole sector whose number of laborers is fixed (it could be a small sector with a unique skill set which doesn’t change with productivity or the whole economy’s productivity increases and number of laborers doesn’t increase), and additionally demand is elastic (i.e. a n% price drop leads to a more than n% increase in quantity demanded), and there is minimal capital required, it works. Say there are 50 cars currently demanded per year and 50 auto workers. It used to take 1 full-time worker a year to produce a car, which costs 20k which is his salary, then productivity increases so he can produce 2 cars a year. If the price of a car drops to 10k then the demand would increase to over 100, but the 50 workers cannot make over 100 cars, hence the price doesn’t drop to 10k but is above 10k, so each worker’s salary is now above 20k.

    • Guy in TN says:

      Increased productivity doesn’t necessarily to drive up the median worker’s wage. If 100 people saw their wages drop by a combined sum of $100,000, but the factory owner saw his wages+capital returns increase by $100,001, then that factory became more productive and saw total wealth increase.

    • benwave says:

      Gripe – a Marxist (me) would say that increasing productivity increases the amount of value produced by the factory in your example. They would not claim that this should necessarily drive factory workers’ wages up though. They would claim that the amount of the extra value that is captured by workers compared to that captured by owners would be determined by relative power. For example, market power by workers being offered higher wages by competitors, or bargaining power by unions negotiating better wages with owners.

      • Swami says:

        I think a free marketer would argue that over the longer haul and broader market (which is now global) that market forces of supply and demand would determine the relative gains. Assuming relatively free entry and exit, the return on entrepreneurial activity and investment would be attracted toward the cost of capital adjusted for risk.

        Thus the explanation for a reduced share of productivity gains going to median wages in developed nations indicates a signal and an incentive that there is not enough capital and not enough skilled labor and that there is an excess of lower skilled labor. The market is saying we need more entrepreneurs (we will pay!) and more skilled labor (doctors, executives, engineers, etc).

        I don’t think it is a coincident that developing nation wages are going up faster than ever before at the same time unskilled labor is stagnating and skilled labor and returns on capital are increasing. The supply of unskilled labor globally shot up faster in the last 40 years than any time in the history of economics.

        And my understanding of the issue is that this is true even in developed nations dominated by unions. Indeed, it is probably more true there, as it has resulted in higher rates of unemployment.

        • benwave says:

          Under conditions of greatly more market power held by owners than by workers, that is what I would expect, yes.

      • Charles Engelke says:

        In the real world, a rise in productivity by a factor of 5 does NOT in fact lead to making 5 times as much money by selling the 5 times more numerous widgets at the same price and paying the worker 5 times more!

        Marxists don’t believe this myth being propagated by Baumol about productivity and neither do actual business people (who have to continually cut cost per unit to stay ahead of their competitors).

        For one thing there is a law of supply and demand (also operative in the pricing of jobs, but I’m referring here to the widgets). Very likely the market for these widgets, at this price, has been saturated. You won’t sell more, those who can afford them at this price are already buying them.

        You need to increase market share, or create a new market sector with the less affluent consumers who previously were unable to afford widgets (model T’s etc).

        Even if manufacturers didn’t want to decrease the sale price;
        Once 5 times as many widgets are sent to the shelves the glut causes store owner to reduce prices just to clear them.

        If, after lowering prices, by say 1/2 and gaining 20 times as many customers, a kindly factory owner wants to give a lot of the new profit to the workers (even if they are not now doing anything physically different in operating machinery) that can last only until the new productivity method spreads, then the competition will hire similar grade workers that would be happy to operate machinery at the old salaries. and undercut the benevolent factory.

        During the boom of an expanding market reaching into new demographics,increased demand for the machine workers will somewhat raise the manufacturing wage relative to average wage, just through competitive economic forces (this happened in mid 20th century).

        BUT, in the long run, such a 5 fold productivity increase will always gravitate toward selling the items at close to 1/5 the original price while paying the worker about the same he has always been willing to receive for this grade and difficulty of job.

        • baconbits9 says:

          BUT, in the long run, such a 5 fold productivity increase will always gravitate toward selling the items at close to 1/5 the original price while paying the worker about the same he has always been willing to receive for this grade and difficulty of job.

          The worker’s real wages are still 5x as high in terms of widgits. If every sector of the economy quintuples its productivity then every worker earns 5x as much.

          • Charles Engelke says:

            Not in terms of work hours expended to buy things which is much more physically real than imaginary increases in money spread diffusely through an economy. I call that kind of rise in nominal money inflation
            (please search down for my posts below with a link to an open document showing how this works)

          • Charles Engelke says:

            I suppose you mean if every sector’s output in fact rises by 5, so not quite as magical as I implied, but the real CPI situation tries to average several sectors that are getting cheaper in terms of work and share of GDP and then renormalize everything (and say there is ‘more‘ money delocalized throughout the economy) which is including things that haven’t changed in output at all (like services).
            A renormalization like that is a human trick with numbers, not a physically real happening.

            It is eerie for someone from the physical sciences to read the conversations based on this, they seem medieval like minutely parsed debates about how many angels on pins supported by and only referencing traditional authority.

            It is doubly bizarre in my eyes since personally I believe the unwarranted renormalization is the source of the mysterious behavior of the angels you are all parsing.

            Do you all know about the “experience curve” beloved in business strategy planning, a declining power-law which declines with a different power in each specific factor or industry, but goes on predicting cost seemingly forever once you’ve determine the correct costant (each unit output becomes lower in cost by a uniform percent (say 20%) for every doubling of accumulalted output. These are real local physical effects. Business men rely on. It’s due to continuous improvement of the physical process. There is no correct weighting that describes a new amount of money in the system. Its different relative to each process.

            There are also very different modes of productivity increase, all physically based. The violinist brought up as an example of something that has no productivity increase can get a doozy of one if he makes recordings that are popular. Same for actors becoming movie stars. Their income does go up directly with the higher number of viewers, though the viewers again pay less than for going to a live concert.

            This is the actual history of our world. We can all read about it if we diddn’t live parts of it .How come the dancing angels?!

          • Charles Engelke says:

            Not in terms of work hours expended to buy things, which is much more physically real than imaginary increases in money spread diffusely through an economy. I call that kind of rise in nominal money inflation
            (please search down for my posts below with a here is link to an open document showing how this works)

          • Charles Engelke says:

            The Bureau of Labor Statistics even tries to account mathematically for the increasing quality of items in their basket when they decide what value to renormalize all our money to.. Supposedly accounting for added experience of increase in standard of living. Sheesh!

            Such intangibles can only be assessed from within (from how the person ‘feels’ their assessment of value — that’s what is “tangible” things that are ‘felt‘, even the desire for a ‘pet rock’)

            The only measure of allowing comparison of how people value different items through time is by the amount of work effort & time they will sacrifice to obtain it.
            And this measure includes things like relative scarcity, novelty (early iphone vs later routine ones) fashion, etc.

            People currently value gold pretty highly by this measure, BUT if after the asteroid Psyche is mined the amount of gold on earth goes up by a factor of 10,000 it won’t even be picked up on the side of the road except by children who like its shine.
            It won’t matter what the BLS woud calculate about the ‘quality’ of the gold!

          • Charles Engelke says:

            baconbits9 said,

            The worker’s real wages are still 5x as high in terms of widgits. If every sector of the economy quintuples its productivity then every worker earns 5x as much.

            To be clear, when I replied, “not in terms of hours expended to buy things”
            I mean both the factory worker and the buyer expend only 1/5 as much of their time budget on, respectively, making the widget, and earning the money to buy the widget. Therefore the widget sector, when the market is saturated, will occupy only 1/5 as much of GDP

            This doesn’t mean GDP has grown
            It means that in very real, absolute, terms the widget has become cheaper

            If a Star Trek replicator just materializes a widget with no work involved, the widget is free, no one has their salary suddenly rescaled to infinity

    • Faza (TCM) says:

      It’s worth remembering that the labour market is also competitive.

      There are several ways this could go down, so here’s just the most intuitive one.

      When worker productivity increases, an employer can lower their wage overheads, whilst still maintaining the same level of production, simply by hiring a smaller number of only the most productive workers (in practice, they’ll probably be looking to increase production – and, presumably, revenue – but we can ignore this). Even if they have to pay higher wages to the individual worker, the total wages paid may be lower, because there are fewer employees.

      Why can’t the employer retain just the most productive workers without increasing wages? Because all of their competitors are trying to do same thing.

      Demand for the most productive workers is easy enough to understand, but what about the less productive? The same mechanism applies: if you cannot get the best (because someone else already snatched them up), you get the next best – but you’re going to be competing against other employers for those, as well.

      I should probably note that the end result may include fewer people working in the industry overall, because demand for its products does not justify more output and – in an increasingly productive industry – a given level of output can be achieved with fewer people.

    • Eric Rall says:

      One piece is that labor productivity, in the long term, puts a ceiling on wages. If you pay your workers more than the value they produce, then you’re not going to be able to stay in business very long.

      Another piece is the effect Faza described, where more productivity attracts competition, which bids up prices.

      The last major piece is that increased productivity across the economy as a whole is deflationary: if the amount of money in circulation is constant and the financial system circulates the money at the same rate, then more goods in the economy means lower prices because there are only so many dollars to go around. So even if nominal wages stay the same, the same wages will buy more goods. In practice, central banks tend to offset the deflationary effect of increased productivity (and then some) by issuing more money, which drives up both nominal prices and nominal wages. So instead of nominal wages growing by 0% and nominal prices falling by 1%, nominal wages grow by 3% and nominal prices grow by 2%.

    • Charles Engelke says:

      For my money, Sarth supplied the wisest “real world” assessment I’ve seen in the whole thread!

  10. blumenko says:

    The Last Psychiatrist’s statement is an example of my bete noir, double counting inflation. Doctors salaries are only stagnant in inflation-adjusted terms, not real ones, so you can’t go on to say that the cost of living has gone up, because that was included in the calculation of the real value of doctor’s salaries.

    • Nornagest says:

      Only insofar as what you’re counting as cost of living is reflected accurately in inflation indices. The Consumer Price Index, the most common inflation measure in the States, is quite broad and includes such things as rental units and medical care, but I haven’t been able to verify that their weighting is close to what an average consumer might proportionally expect to pay.

      • blumenko says:

        So every blogger is allowed to claim that they know better than the government statisticians whose job it is to compile the data?! I claim visibility bias; the same reason people care more about gas price increases that have way less impact on their income than other subtle price increases.

        • zzzzort says:

          I think there’s a coherent (and not too terribly postdictive) argument that the CPI, which weights costs on something like the average consumption basket, undercounts inflation faced by bougie people who disproportionately spend college and medical care.

        • Nornagest says:

          Why yes, in fact, every blogger is allowed to claim that. I’d prefer some effort invested into untangling how it works in nuts-and-bolts terms (as Nybbler did below), but it’s perfectly acceptable to claim based on anything down to and including vague personal gnosis that the official measures don’t adequately capture cost of living. An unsubstantiated fetish for government expertise is just as bad as any other argumentative sin.

      • The Nybbler says:

        Current weighting is about 33.2% shelter, split between 7.9% actual rent and 24% “Owners’ equivalent rent of residences”, 7% medical care (1.7% physicians, 2.3% hospital), 20% commodities less food and energy, 6% transportation, 7.8% energy, 13.2% food.

        That report shows shelter has increased 3.3% from May 2018 to May 2019, medical care 2.8%, transportation up 1.1%. Commodities less food and energy down 0.2%, food up 2%, energy down 0.5%, netting out to 1.8% for all items (2% all items less food and energy).

        • Nornagest says:

          Thanks. Commodities sound high and transportation sounds low to me (even considering that gas is probably folded into “energy”), but this is probably very sensitive to class and geography among other things.

  11. Guy in TN says:

    Two legitimate questions I don’t know the answer to:

    How much would change if you switched from looking at the median wage to the average wage? Since the Baumol effect relies on an increase in total demand this is more accurate, no?

    And how much would change if you measured the increased wealth from labor+capital instead of just labor? For the same reasons as above.

    • blumenko says:

      Baumol effect is a supply side, not demand side explanation. Increased demand for a fixed supply is an alternative explanation, but raises the question as to why supply is fixed.

      • Guy in TN says:

        In the case of the violinist who now has the option of changing to a lucrative career in basically anything else, does this not mean that the demand for warm bodies has increased? Otherwise how is the violinist able to command a higher salary?

        • blumenko says:

          It doesn’t necessarily. If demand hasn’t increased there will be fewer violinists, who were charge a higher price. If demand has increased, as it usually will if everyone gets richer, then there will be the same number of violinists, at a higher salary (this is assuming that everyone who could become e.g. a computer programmer could become a violinist, complete fungibility of labor in the most strict version of the Baumol effect, if we assume that there are somewhat different talents involved in being a programmer and being a violinist, then demand will have an effect on price, but due to lower supply the price will always be higher).

    • Ketil says:

      How much would change if you switched from looking at the median wage to the average wage? Since the Baumol effect relies on an increase in total demand this is more accurate, no?

      +1. If income is very unevenly distributed, the median can easily hide a lot of the total wages. E.g., if 60% of doctors make $100K, and 30% make $200K and 10% make $800K, the average is 200K, or twice the median. Such a distribution is probably more true for college-educated professionals, so I would expect the graphs to diverge even more.

    • 10240 says:

      Indeed. Increased company revenues due to increased productivity do go somewhere. According to Scott’s data, they mostly go to high-skilled professionals’ and managers’ salaries. The Baumol effect can work on managers’ salaries too: if managers’ salaries have increased in high-growth sectors, then managers’ salaries in no-growth sectors will increase too.

      A counter-argument is that government organizations like schools don’t tend to hire very-high-salary managers. Another counter-argument is that companies hire the best managers for millions of dollars because they are the best at squeezing out the maximum productivity increase, which is why they reap much of the productivity increase. If it’s impossible to increase productivity in some sectors anyway, they could just hire mediocre managers who keep the shop running without changing much. However, medicine is not unchanging: if, say, new technologies and practices improve outcomes of some procedures, that may not show up in the official productivity statistics. But if you didn’t adopt the new technologies, you would be providing substandard care, and if the changes are managed by mediocre managers, that can lead to inefficiency.

      Capital matters less: capital’s share of the GDP has only slightly increased, and it has slightly decreased if housing is not counted.

      • 10240 says:

        Capital matters less: capital’s share of the GDP has only slightly increased, and it has slightly decreased if housing is not counted.

        This part was inaccurate. Profits as a share of the GDP haven’t changed much, but the GDP has increased (per capita, adjusted for the CPI). However, it seems to me that rates of return on capital have decreased; the increased amount of profit reflects an increased amount of capital (all adjusted for population and CPI). On the other hand, the prices of capital goods may have increased faster than the prices of consumer goods, so it may take more capital (on CPI-adjusted dollar terms) to run a business, even in an unchanged way, than in the past. I don’t know how the profits on a given amount of capital goods have changed compared to consumer prices (that is, (rate of return)*(capital good prices)/(consumer prices)).

  12. Toggle says:

    (For those who want to read Tabarrok. Also, gentle finger waggle at Scott for not linking to it.)

    • Clutzy says:

      Not going to say he’s crazy from the start, but these sorts of statements always send into a tailspin of citations needed:

      As an empirical economist, I am interested in testable hypotheses. A testable hypothesis is that the industries with the biggest increases in regulation have seen the biggest increases in prices over time. Yet, when we test that hypothesis as best we can it appears to be false.

      How do you measure this? One of worst possible ways would be number of regulations. His cars example makes him look extra silly here. Car regulations have basically eliminated entire departments from companies, because they can’t compete on design anymore. Cafe and other regulations just set specific targets in areas customers always wanted (like fuel efficiency and safety). The major cost increase in automobiles is that customers can’t get the cars they want anymore because entire lines were eliminated by regulation. Like Minivans and station wagons from certain companies.

      The best way I ever heard this said by was some cooky late night radio host in 2010ish who was talking about executive orders. He made the point that the number is pointless, because a President could issue an executive order to every agency dictating the color of pens they have to use. Indeed he could issue one to every executive employee. That would be millions of executive orders, the sum of which are less important than one order on one interpretation of one line of the tax code. Plus, a President could also interpret all the ambiguous lines in one order, which would out-value all of the executive orders of all the presidents since 1900 combined.

      • RalMirrorAd says:

        The most reasonable approach seems to be compliance cost, but measuring that accurately is no small feat. However that approach still doesn’t account for survivorship bias and the potential leftward shift of the supply curve [from market exits]

        There are other proxies used in measures of economic freedom like the number of days it takes to get a business permit.

      • Garrett says:

        In the case of medical care costs, how much of this is based on the change in “the standard of care”.

        When you look at things like violin concerts, we’ve gone from 2 significant options: no concert, or a live performance you pay to attend, to 3 significant options: no concert, a live performance you pay to attend, or a recording of such a concert you purchase once and can play on demand. In this case, you might argue that a recording is an inferior good (or superior, along different dimensions). That there are more options allows more consumers to satisfy their preferences more optimally.

        With things like medical care, providers cannot legally (it’s complicated) provide a lower quality of care than “the standard of care”. Indeed, if a doctor was to offer only the 1960s standard of care today (except, perhaps, when current standards are better along all possible dimensions), they’d likely lose their medical license and/or be sued into oblivion. Indeed, because the rest of the whole industry is organized around the current standard of care, there would be minimal cost savings to doing so. Only by completely disbanding that concept would a wide range of products and services be able to open up and allow for optional savings and competition.

        Imagine trying to get something through a FDA regulatory approval using the 1960s standards today.

    • Scott Alexander says:

      Thanks for reminding me; I’ve added the link.

    • Ben Wōden says:

      From the Tabarrok piece linked: “Bryan, the armchair economist…”

      I rate this level of inter-academic saltiness at the unprecedented level of 1 micro-Taleb

  13. TheTime says:

    So… can we have a Baumol effect of capital productivity instead of labor productivity in out hands? I have no idea if it makes sense, but it sure seem like machines help people be more productive A LOT nowadays, which seem like it could potentially make capital more expensive even when and where it is not very useful –
    If I’m selling computers for administrative purposes, I can sell them to big successful companies or I can sell them to schools. The cost of my computer presumably determined by the increased productivity of the administrators. Let’s assume that the school productivity is relatively constant, and that the companies’ productivity is going up. As the schools actually need the computers to administer stuff, they need to buy computers ever if their cost have gone up thanks to productivity gains of bushiness.

    Can something like that explain the raising cost of education infrastructure?

    • Guy in TN says:

      You’re basically right, but I wouldn’t draw a bright line between capital and labor productivity in this case. You could see the Baumol effect on wages if the productivity of capital alone rises.

      For instance, a more productive factory would allow someone to purchase more violinist-hours. So hiring violinists becomes more expensive, despite violinists not being any more productive, because demand for violinists increases but the supply stays the same. Classic Baumol.

      This is why any analysis that looks solely at labor will be inadequate.

      • 10240 says:

        Your argument is not exactly the same as the Baumol effect. Your argument looks at increased demand, while the Baumol effect looks at the violinist’s choices. If much of the benefit of increased productivity goes to labor, a violinist could become a factory worker (or manager) instead, so it pushes his wage up. However, if most of the benefit of increased productivity went to capital (which would be the case if the supply of capital was very limited), a violinist couldn’t choose to become a capitalist instead: he may not have much money, and even if he has money to invest, that doesn’t preclude also playing the violin. The number of people with enough capital to not have to work is too small to significantly affect the labor market.

    • Garrett says:

      The flip-side is that lower costs for administration make it a lot more justifiable to add a lot more administrative tasks. It’s possible that the reduction in the cost of an administrative task has resulted overall in more time being spent on administrative tasks overall because the incremental cost is so low.

  14. Radu Floricica says:

    Suppose that in 1960, widgets cost $1, a worker could produce ten widgets an hour (and made $10 in wages)

    This rubbed me the wrong way very hard, and I might have stumbled onto something here. In real, contemporary life you have to make closer to 100 widgets an hour to actually get paid $10, even for a self-employed. There are supplies to be paid for, service providers, income taxes, sales tax, part of the profit has to go to a bad times buffer – long story short, you just don’t get paid even close to what you produce. But isn’t that a damn good explanation for what we’re looking at? 70 years ago you may have been able to walk in from the street, work a week and get a week’s wages – not any more. It may have been possible to buy raw materials, do stuff to them, sell them, move on. I can’t even imagine how this would look today and be even remotely legal.

    A couple of my friends are thinking of closing a cafe that’s very dear to me. I tried doing the math on how it would cost to keep it barely open. The real, material extra cost would be what you need to pay an employee to keep the lights on every evening, not even full time. Considering tips and that’s it’s a popular cafe in a town full of unemployed young people, the actual sum would be trivial. And then I went on to the calculate the actual, real-life amount, and it kept increasing and increasing and…

    There’s also a more speculative point of which I’m unsure, but while thinking about the cafe’s expenses I started double-counting Value Added Tax. In theory a business only pays its share once, that’s its whole point. But the building’s owner is a private person who’s not including VAT in his rent, but who’s using the rent to pay for his car that definitely has VAT. I’m a bit confused on this point, but it might lead to something.

    This shouldn’t happen in real life, because if you work in the high-productivity-gain industries, you should make more because of your increased productivity, and if you work in the low-productivity-gain industries, you should make more because of the Baumol effect

    Quick note that Baumol effect is dragging wages up the hard way. It’s very much not a painless process – you get less farmers because they get bankrupt, or move to something else, or retire and nobody goes into an unprofitable industry anymore, and as a result the wages for the ones that stuck get up barely enough to create an equilibrium. Or simply because the state mandates that there are minimum wages, which does pay them minimum wages but has the sword of a failed business hanging forever over their heads.

  15. Nornagest says:

    I play with martial arts equipment and make smart remarks on the Internet on my days off; Scott picks fights with PhDs in other fields. It’s days like this that I feel a little bit inadequate.

  16. blumenko says:

    I think there is a bit of an overexaggeration in your critique re: share of pie increases vs. absolute increases. If salaries command a large majority of education costs, as they do, and their share of the education pie has roughly stayed the same, as it has, then increase in salaries is the cause of the majority of the rise of education costs, and explaining that explains most of the issue. It would still be unexplained why other costs are rising, but that would be a minor issue. One could point to that and say there is a common cause to both costs rising, but that may not necessarily be true.

    • DarkTigger says:

      It would explain why the other costs are rising. Everyone needs education, education gets more expensive, people need more money to pay for their education, so they need to rise prices.

      Or do I miss anything?

      • blumenko says:

        I mean other education costs, not other sectors of the economy. We are talking here about prices rising faster than general inflation, and general prices can’t rise faster than general inflation.

        • DarkTigger says:

          I mean other education costs, not other sectors of the economy.

          But the other costs are determined by other parts of the society, aren’t they?

          “Benefits”, like health insurance are bound to health care. Which also is affected by the Baumol effect.
          “Services” are bound to industries like cleaning and transport, which probably also haven’t seen big productivity boosts in the last 50 years. So are also affected by the Baumol effect.
          “Supplies” got smaller, and is coincidentally the one thing on the that is not affected by this effect.
          So if most of your costs are affected by the Baumol effect, but don’t profit from falling prices in production, you will get effected harder.

          I mean I am just spitballing here. I’m happy to be shown that I am wrong.

      • A Definite Beta Guy says:

        No, because other sectors of the economy may have productivity growth. True, labor may be more expensive, but I may have labor-saving technology that allows me to produce more with less labor.

        In contrast, teaching is not getting more productive, and is labor intensive. So any increase in labor costs will increase the total cost of the good.

        In addition:
        1. “Education” is broader than it used to be.
        2. “Teaching” may actually be less productive. Class sizes are smaller, and teachers have more support staff in the schools.
        3. “Education” is influenced by political choices, not just market choices.


        Everyone needs education

        Do they, though? Do they? Can we start stripping things OUT of education that aren’t really needed, or change the model to something less labor intensive so it becomes less expensive?

        Anyways, I had the same objection as Blumenko: if labor is holding constant at 80% of the pie, and the pie just increased by 5%, then labor also increased by 5%, and is the biggest component of the pie increase.

  17. blumenko says:

    I still think that the Medicare cap on residency funding is a red herring in terms of overall healthcare spending. Hospitals are free to have more residents, they just won’t be subsidized by the federal government. So if demand really justifies more doctors they would be trained at no more cost to society than currently. It would just shift more of the cost to the private sector, and increase both doctors’ salaries and medical school debt relative to the current situation.

  18. benf says:

    Salaries in service sectors not rising is not evidence against Baumol’s cost disease. It’s evidence of low inflation and/or low labor bargaining power. The relevant numbers are shares of the workforce being employed in service sectors versus manufacturing/agriculture. The story is, manufacturers invest in efficiency, make more money, use that money to employ the labor they used to employ in their factories to cut their hair or teach them English or whatever. If manufacturers all decide they want haircuts, salaries for barbers may rise a little bit, but the higher price will lead to an increase in the supply of barbers and the price will fall again. So barber salaries stay the same but now there are more people cutting more hair instead of operating lathes, because the lathes are running themselves. This is how automation and capital investment change the economy and increase standards of living overall, but not necessarily for everyone individually. It’s also why automation will not make humans obsolete: there will always be something you can find for a human to do if you have the income to spend, even if it’s just paying them to smile at you or do a job that machines can already do perfectly well, like making cappuccinos.

    • AlesZiegler says:

      Yup. I have essentially the same point but you were first.

    • Aapje says:


      Your argument is wrong due to multiple reasons, including:

      1. People don’t necessarily have sufficient talent to do whatever pays, no matter how high the salary.

      2. Most people don’t spend just to spend, but to get their needs met. However, even then they balance sacrifices they make for their income to their spending. They are loath to spend on things that cost them less personal sacrifice to do, than the sacrifice needed to earn the money to pay someone else to do it. Yet:
      2a. Income/hour tends to be discontinuous, so people are often forced to work more or fewer hours than they would prefer. If they work more hours than they prefer, they thus make more sacrifices for part of their labor than they would like. This in turn causes a market distortion, where they are loath to spend their money unless it brings them greater benefits than marginal sacrifice of their time.

      A simplified example: Bob would like to work 20 hours a week for income $500 at a sacrifice of 50 utils, but his actual options are to work 40 hours for $1000 at a sacrifice of 70 utils or 20 hours for $300 at a sacrifice of 50 utils. So he works 40 hours, grudgingly.

      Yet his preferred spending pattern is to spend $500 a week to gain 200 utils, where spending that other $500 only gives 10 utils. Yet because that extra $500 costs him 20 utils (70 – 50), this feels like a losing proposition. So he is incentivized to either save that $500 or to spend more than $500 (with a loan) to gain at least 20 utils from the spending beyond $500. Either feels fair, while spending that $500 for so few utils doesn’t feel fair.

      Currently, the US tends to pick the loan + spend option, while my country tends much more to savings. The former is unsustainable, while the latter results in unspent income.

      there will always be something you can find for a human to do if you have the income to spend, even if it’s just paying them to smile at you or do a job that machines can already do perfectly well, like making cappuccinos.

      Except that technology allows relatively few people to service very many people in this manner. See camgirls and YouTube stars.

      Furthermore, some technology is actually better overall than humans. Why did local bands mostly disappear from cafes? Because most local bands don’t have the talent to compete with a recording of a very talented artist. The local artists’ talents became worthless, as they were replaced by CDs.

      The risk of automation is that most people simply can no longer compete: people are ‘friends’ with PewDiePie, who is more interesting than their neighbor Bill. They have ‘sex’ with a (VR) porn star, rather than their not so pretty neighbor Mary. They get their coffee from a robot who is more pleasant to be around than Bill or Mary and who actually produces better coffee. The cleaning lady is a Roomba.

      Then Bill and Mary are obsolete as workers. They have nothing to provide to Richy McRich that he can’t get better from a few other Richies.

      • carvenvisage says:

        @aapje just a tip for wording as I know you aren’t a native english speaker: “worthless” has very negative connotations as a word; to say something is worthless means not so much “without utility/economic-value” (*the speaker adjusts his monocle*) as “it’s a piece of shit”.


        Why this is; I think partially it’s just inertia- the word gets used that way often enough and it becomes a word to use that way, but there is also some logic to it:

        1. “worth” is a synonym of value, which is a synonym of utility, but the former suggests a personal assignation of value much the same way that utility suggests a clinical analysis of it. Observe that “self worth” has listed synonyms of self-respect and self-esteem.

        2. “-less” as a suffix suggests not just an absence but a lack.

  19. AlesZiegler says:

    Unless I am badly missing something, second objection could have fairly simple explanation. Wages of doctors/teachers/violinists have not risen because there is more of them. Share of GDP consisting of education/medicine/classical music increased.

    Now, it is empirical question whether this happened, which at least in US healthcare probably could be answered by resounding yes.

    • sourcreamus says:

      Yes, they make the point that doctors per capita have tripled in the time period. So although individual compensation has not changed that much, compensation to doctors as a group has skyrocketed.

  20. tmk says:

    My problem with the first two graphs, and I think it’s a serious one is: They are looking at slices of the population without considering the changing size of those slices. Salaries may be kind of stagnant for both high school only and college grads, but a much higher portion of the population are going to college!

    That also means the general quality of both groups are lower. Back in the days a lot of people only did high school (or less). Today it’s, to be very crude, only the people are two dumb to get into college. But they are still getting paid about the same as high school grads from back in the day. That is a rising wage per work quality. Likewise, a big portion of 2019 college grads aren’t as bright at 1960 college grads, but they’re still getting paid more.

    Teachers salaries are held down artificially for political reasons, but that just means the market reacts by attracting lower ability people to teaching. That means you get less work value out of each teacher, even if the number of pure teaching hours are the same. If I’m allowed to make things up: In 1960 you could expect one teacher to pretty much run a small rural school by themselves. Today anyone with all those abilities have better paying options, so you have to hire administrators to support your teacher.

    • Simon_Jester says:

      There are a few complications about the nature of modern education that aren’t in this analysis.

      Firstly, “standard of care” has increased markedly in high school education since the 1960s. This is relevant for a few reasons:

      Teachers salaries are held down artificially for political reasons, but that just means the market reacts by attracting lower ability people to teaching. That means you get less work value out of each teacher, even if the number of pure teaching hours are the same. If I’m allowed to make things up: In 1960 you could expect one teacher to pretty much run a small rural school by themselves. Today anyone with all those abilities have better paying options, so you have to hire administrators to support your teacher.

      In 1960, there was much less pressure on said teachers to graduate the bottom 10 to 20 percent of a given school’s student body. Not least because a person could lead what was, by the standards of the time, a normal life without being a high school graduate. This has become harder today, and by the same token teachers are expected to be far more accommodating in terms of what sort of student behavior and background knowledge are even allowed to enter their classrooms.

      A teacher running a small rural school by herself could unilaterally expel the kid with ADHD as “disruptive and uneducable” or what have you. She could take the girl who spent the whole tenth grade staring out the window despite her best efforts and make her repeat it. And it was far less likely that people would try to gainsay her in this.

      This kind of thing makes the teacher’s job vastly simpler.

      If you can say “I won’t be having with this, nobody passes 11th grade English unless I’m satisfied they understood it,” and nobody minds, you have a lot more leverage to teach a rigorous English curriculum. Or to teach an English curriculum no harder than one from the present day, but to work a lot less hard in the process of doing it.

      The result of this is that we’ve made teaching a more specialized profession, at least if we’re going by the standard of “so, how good to you have to be in order to do this?” Just as a doctor who provides a 1960 standard of care in 2020 is likely to get sued into the ground rather than offering customers a low-cost alternative that saves money, a teacher who relies on the toolkit of 1960 to teach the classroom of 2020 is going to be regarded as incompetent and very possibly hounded out of the profession.

      This in turn necessitates support staff. The school didn’t used to need a school psychiatrist or crisis counselors. Or, from a different point of view, it DID need them, it just didn’t have them and sort of shrugged when mentally ill students or those having trouble at home just dropped off the radar and never came back. The school didn’t used to need security guards so badly because the kids deemed likeliest to need them were pre-emptively expelled, sacrificing their chance to shape up and sort themselves out in exchange for making it easier to educate everyone else.

      In other words, the consequence of making nearly everyone a high school graduate isn’t just a decline in the intelligence of the mean and median high school graduate. It’s increased workload, and need for support staffs, for the people responsible of getting those kids through high school. Because we didn’t bring about ALL the change in graduation rates just by lowering standards; some of it came from raising our game.

      • glorkvorn says:

        I was thinking along the same lines as you. It seems apparent to me that the growth in “teachers” is not in the sort of traditional teacher you might imagine (standing in front of 20-30 kids teaching reading, writing, and arithmetic). It’s in special ed teachers, particularly for ESL or children with disabilities. They have to be taught invidually or in very small groups, so it’s very expensive, and it’s probably not going to boost test scores very much, but special ed is still important. Plus all the administrators to oversee them.

        Healthcare, similar story. It’s not really about traditional doctors and nurses, but in niche roles like medical technicians, often focusing on a specific technology or type of patient. Every time some new medical technology is invented, we need a support staff for it, but we can’t get rid of the old support staff either. I can’t decide if this is a good thing (an expensive treatment is much better than no treatment at all) or a bad thing (our society is throwing away all it’s resources on desperate measures because no one can say “no”, like some ancient Chinese emperor seeking the elixir of immortality).

        • Matt M says:

          but special ed is still important.

          At the risk of being insensitive, uh, citation needed?

          I’d be curious as to what sort of societal cost/benefit formula you might run that would project a positive return for 12 years of forced education for those with severe learning disabilities.

          • Dack says:

            A thing doesn’t have to be profitable to be important.

          • Andrew says:

            While that’s true, the reality is that trade offs happen whether we’d like them to or not. I have a friend working in the school system we both came up through 2 decades ago. Through casual conversation, I gather special education spending has gone way, way up, and the math & science magnet school I attended and largely credit with my not crash & burning in college was shut down.

            I don’t have anywhere near enough information to tell if that was a good trade, but even if we accept that special ed is important, other things are important too.

          • thisheavenlyconjugation says:

            Yes, if you run the numbers it turns out that processing those with severe learning disabilities into a thin gruel that can be used to feed more productive members of society would be more optimal.

    • blumenko says:

      Teacher prices are held up artificially for political reasons. Teacher’s unions are huge contributors to political campaigns, and if the prices would be held down artificially, then the unions would be all for charter schools, which would increase demand for teacher’s salary. The fact that they oppose them, means they are being paid above-market rates.

      • Edward Scizorhands says:

        Different states have different teachers’ unions. Many states have “unions” which cannot strike or enforce membership, so they cannot reasonably be holding salaries up.

        Do we see differences in support for charter schools in those different states?

  21. HeelBearCub says:

    It seems to me that a mistake is being made here.

    Scott, you are alternating talking about prices and wages. Baumol isn’t about wages, it is about prices. Increased productivity increases supply, which (Econ-101) lowers prices. Any wages are downstream of this effect. Low productivity gains for one product means that prices rise in relation to the high productivity gain product.

    For an example of how this can trip one up, consider this post from Kevin Drum about a proposed switch to chained-CPI for measuring the inflation rate. The key point here is that “hedonic adjustment” is a key component of inflation calculations.

    Theoretically, the real cost of a car has halved since 1980. But when we look at the inflation adjusted cost of a lowest base price car today vs. in the past, they are basically flat (not halved). But that’s because the productive gains involved in making a car haven’t gone into lowering the price of the out-the-door car, but rather in making “more” car (ABS, air bags, power windows and doors, etc.).

    The professor, doctor or nurse’s productivity isn’t rising at the same pace, so relative to the flat price of the base model car, the “base model” healthcare increases. There’s more car available compared to healthcare or education.

    The assumed argument is that we should should somehow see this directly reflected in wages. It’s probably an interesting argument to have, but it seems to me it’s a different one.

    • Conrad Honcho says:

      Yes, this is definitely something I have trouble understanding when it comes to comparing prices and wages from different eras: how exactly do we make sure we’re comparing apples to apples? You can’t even just compare apples, because automation, pesticides, logistics, etc, allow fewer people to more efficiently deliver more apples to more people. And you can’t compare what you could trade those old apples for versus what you trade the new apples for because, like you said, however many apples got you a car in 1980 might get you a car but a completely different quality of car in 2019. We’re not doing the same labor, and we’re not producing the same things.

      • Aapje says:

        But it goes the other way too. Some things, like having a servant, have become more expensive. To what extent is the shift to alternatives a deterioration that people feel forced into due to rising costs of certain goods/services?

      • HeelBearCub says:

        The short answer is that you track inflation, not the price of apples.

        And you track inflation by measuring a whole “basket” of goods that represent the broad economy. There are different ways of doing this, and those different methods come up with subtly different numbers, but they don’t differ wildly from each other.

        Hedonic adjustment is included in the calculation of inflation. But that means that sometimes inflation doesn’t capture what prices “feel” like. Remember that, according to inflation calculations, cars have dropped in price (in real dollars).

        But it doesn’t feel that way because the average person buying their first car today is buying more car. A whole, whole, whole lot more car.

        Or, take something like phone service. From a real dollar perspective people are paying way more for “phone service” today. But from a hedonic perspective they are paying way less!

        • Conrad Honcho says:

          Yes, and my intuition with this is “basically you can prove anything you want about the economy with numbers by looking at it from a different perspective.”

          “Are people working for more or for less than they used to?”


          • HeelBearCub says:

            Just because “90% of all statistics are made up”, it doesn’t follow that statistics are meaningless, though.

    • blumenko says:

      Every explanation I have seen is that the Baumol effect is about wages, not prices. Your posited effect doesn’t make sense without a wage picture. The input to a string quartet isn’t really 4 string players, but the wages needed to pay 4 string players, which covers housing, food, mobility etc. If the cost of all those goes down, then the wages (and cost) for the string players should go down as well, as that would keep the real wage the same. The reason it doesn’t because of rising real wages.

      • HeelBearCub says:

        Hmmm. I see Wikipedia seems to agree with you. I had always seen it being referred to as “Baumol’s cost disease” and people complaining about the cost of all sorts of things going up.

        Thinking further, I believe I am inclined to agree that the basic argument of Baumol works at both the wage and the price level.

        However, I still think it’s a mistake to think of wages and prices as fixed relative to each other. Just because these effects result in price increases, it does not necessarily follow that we will see an equivalent rise in wages. It is a mistake to talk about wages and prices as inextricably linked as if in a fixed formula. We know this isn’t the case.

        • blumenko says:

          People are complaining about rising prices and Baumol’s disease is a particular explanation that focuses on wages. Whether it applies to a particular case is determined empirically, so it is not a fixed formula. There may be alternative explanations, but the point is to distinguish between them.

  22. Blueberry pie says:

    Has anyone tried to get data on this outside the US? That might be revealing – since I would guess western countries are likely to have roughly similar Baumol-style effects while they differ much more in regulatory environment and other factors.

    For example here, in Czech Republic, I am quite sure the expenses on health and education are not growing dramatically (could try to dig some sources if anyone’s interested) – part of that is obviously that the state pays for almost all education and health, but it is not markedly increasing as a percent of the budget/GDP/whatever. We are however also having large problems due to government not paying teachers and doctors enough so they leave the profession/move to other countries. Is that something to be predicted if Baumol was the cause of cost disease in the US?

    • Ghillie Dhu says:

      Absolutely. The core of Baumol is that the laborers have an alternative: sectors with increased relative productivity; if compensation in the decreased relative productivity sectors does not keep pace, they’ll switch sectors. The last step (compensation increasing in the decreased relative productivity sectors) is short circuited by government involvement in your example.

  23. bernie638 says:

    I don’t think anyone really has the right tools to analyze economics properly. This is a really big country with really big differences between places. For example, in my zip code the median house costs just under 90K, and the median teacher makes 49K. Using the thumb rule of 3x annual income for a house, a median teacher can afford an above median house. This is probably not true where you live. I don’t think national medians or averages work since they are skewed by more people living in expensive cities than in the rural areas.

    Looking at the same place over time doesn’t fix the problems since places don’t stay the same. You could try to compare like places over time, like 10 largest cities average 1970 to 10 largest cities average today, but it’s not easy to find good data like that.

  24. niohiki says:

    3. The Baumol effect cannot make things genuinely less affordable, but things are genuinely less affordable

    I think I screwed up here.
    So I retract this third objection.

    Sure, there’s no objection about the Baumol effect not making things more expensive on average. But I think that the real reason anyone would be interested in explaining things with the Baumol effect was to say, as per Scott in the previous post,

    And I want to believe. If this is true, it’s the best thing I’ve heard all year. It restores my faith in humanity.

    If we are left with “the Baumol effect leaves an open door to rising inequality that goes directly into No-Idea-How-To-Fix-It-Land” then I anyway don’t have so many reasons to share the optimism of the authors when they say

    Society can afford just as many string quartets as in the past. Indeed, it can afford more because the increase in productivity in other sectors has made society richer. Individuals might not choose to buy more, but that is a choice, not a constraint forced upon them by circumstance.

    (See how happily they switch “Society” and “Individuals” there.)

    • Orion says:

      I’d say that “normal people can’t afford to pay college tuitition or medical bills because capitalists have taken all the money” is slightly more optimistic a take than “normal people can’t afford to pay college tuition or medical bills any more because we forgot how to run schools and hospitals” and much a more optimistic take than “normal people can’t afford to pay college tuition because capitalists took all the money AND we forgot how to run schools.”

  25. pdan says:


    The “sticker price” of college has risen dramatically, but the AVERAGE price has basically kept pace with inflation. We live in a world in which the rich pay the most, and the poor pay the least. This should be celebrated with bells ringing in the streets. Instead, it’s a constant complaint, and I have no idea why.

    • The Nybbler says:

      Perhaps because when we ask the price of something, we want a number, not “How much is it? Well, how much ya got?”

      • Matt M says:

        But what good does that number do when you know that 95% of people won’t be paying that price?

        I think pdan is onto something here. College and health care are both where we see the biggest supposed rising costs, but they are also the markets where the overwhelming majority of people are unlikely to actually pay anything close to the “sticker price.”

      • pdan says:

        We could fund universities publicly, and we’d have largely the same result. The rich people would pay a great deal more in taxes, the poor would pay very little. There’d be no incentive for people to assess their true ability, though, so we’d need to introduce a testing system.

        The only part of this that bothers me is that the people I hear complaining about the high cost of college (biased sample, but still) are those who have benefitted the most.

        We could address the other major shortcoming by tieing the amount of funding a university can get in student loans to their loan repayment rates.

        • bernie638 says:

          I don’t think it’s the same result since (my limited understanding) there are a lot of people from other countries that pay the full sticker price. I’m not sure how you would keep that dynamic in a public funded system.

          • thisheavenlyconjugation says:

            Why couldn’t you? The UK does — a British (and for the moment, EU) medical student will pay ~£9k/year, whereas a non-EU international at e.g. Cambridge would pay £55k.

          • 10240 says:

            You could have non-citizens pay the full cost just the same.

        • blumenko says:

          That is shifting the costs of college from those who benefit the most (college students) to everyone, including those who don’t go to college and don’t benefit. And no outside testing scheme will be as good as charging and thereby weeding out people who it isn’t worth it for. As it stands a rich or poor person who doesn’t plan on sending anyone to college faces no effective marginal tax on earning more, but under your plan they would, and they would reduce output, for a deadweight loss.

    • Edward Scizorhands says:

      Instead, it’s a constant complaint, and I have no idea why.

      Because it sucks for the provider of a service to dig through your life to determine exactly what you can afford to pay.

      It’s yet one more thing on the list of marginal taxes put onto the lower- and middle-class.

      • Matt M says:

        Except, technically speaking, they aren’t doing this in order to raise your price, but rather, to lower it.

        As far as I know, you’re free not to fill out a FAFSA. I think most schools won’t rescind your admission if you don’t disclose your family’s finances. You’ll just be required to pay the full sticker price.

        • Edward Scizorhands says:

          they aren’t doing this in order to raise your price, but rather, to lower it.

          No. They are doing it to charge people the maximum they can.

          They frame it as at giving a discount, but that has no effect on reality. This is the standard for the corporate pricing world.

    • Swami says:

      Thanks for posting this. I made a similar observation on one of the earlier posts. Because of grants,discounts, and tax deductions the actual affordability of college is not rising anywhere near what the sticker prices seem to indicate. And on medical care, I think it is hard to argue that in terms of benefits in terms of outcomes, that we aren’t getting more for our money than we did twenty years ago.

      My health care costs are twenty thousand a year for two people, but I would still rather pay today’s prices for today’s care than 1970 prices for 1970 health care.

    • beleester says:

      I can think of two reasons:
      1. If “financial aid” in that article includes student loans, then you are paying those costs, just later.

      2. If your financial aid comes from the government, whether as a loan or grant, then everyone is paying for it indirectly (although the rich do pay a larger fraction). “The government needs more and more money to give everyone a college education” is not the sort of news that gets people dancing in the street.

  26. Seth B says:

    Here’s a paper about Baumol’s in US post-war history that might address some of your lingering questions:

  27. Charles Engelke says:

    I am providing a link to a google document I made (in response to Scott’s last “Baumol” post) showing some plots of historical data relevant to what I am convinced is the underlying explanation Scott is looking for (without knowing it).

    These include some plots regarding declining median wage vs college costs, but I will have do some more to address doctor salaries and such in the new post. But my first concern is to elucidate what is going on.

    The Baumol effect was intended to explain, using just the interplay of natural economic forces, why all service wages rise in lockstep relative to that remainder of the economy which is using technology to increase output per worker.

    Scott liked the universality of this attempt, but seemed to recognize an absence of any sort of transparent incentive/reward pricing-feedback signaling that could be regulating it.

    Indeed, whatever might have been the case in the 1920’s when mass production was first expanding; at present 90% of US workers are in services while the sectors with improving productivity employ fewer and fewer workers per unit of output in basically saturated markets. Not a recipe for maintaining upward price pressure in order to keep more people in services.

    My belief is that the fundamental cause of much such seeming nonsense in economics is that inflation is being grossly miscalculated by the professionals. The true adjustment should keep service pay (and all comparable average-skilled worker pay) constant per hour worked. Human beings trading physically experienced time and effort (either directly, or through its fruits) provide the unvarying standard measure of the cost (and thus valuation) of things in any time. In the new constant average wage reference frame obtained by adjusted for this “absolute inflation,” the high productivity output goods can be seen to be getting cheaper & cheaper in terms of the hours one need work to pay for them as productivity increases.

    The document is hastily assembled (and still a work in progress) but I would like to know people’s reactions to my plots showing the “actual” appearance of historical economic trends when one scales all prices to be indicative of how long the average wage earner in each era would have to work to pay them.

    At the least, I believe that anybody would find this more of an insight into the way people actually experienced the economy they lived in. How much work-sacrifice does the average new home buyer expend in different times? Apparently, about the same throughout the last century. How about to buy food? It was notably more expensive in the past, for example paying the 2018 work equivalent price of over $20 for a dozen eggs in the 1930’s.

    • nkurz says:

      Nice post, and a good place for it.

      I clicked the link to your Google Doc, and it said that I didn’t have permission to access the document. There was a button to request permission, but is there maybe a setting on your end to make it automatically public?

      • Charles Engelke says:

        Thank you for telling me!
        I think it’s fixed now, try again

      • Charles Engelke says:

        I have now added the salaries of American physicians from 1928 to 2003 from Scott’s doctor link:
        The salaries of doctors have declined in my ‘absolute’ inflation terms!
        Down 40% just since 1996.

        All these worries Scott and the d*mn book talk about are consistent with being artifacts of false inflation adjustment (CPI usually only 60% of what I recommend per year to keep average work per hour constant)
        and as Copernicus would say, mine is the more parsimonious hypothesis

        At the very least I hope people recognize that the way the doctors experienced their income in each era is better indicated by how long the average worker in he same era must labor to pay

        Is Baumol going to claim that doctor’s salaries stayed in step with other services BECAUSE OTHERWISE they would GO TO WORK in a FACTORY!??

        They were earning the equivalent of $550,000 2018 dollars in ratio to average of time, for goodness sake! How can factories make medicine stay in step with the other services over time?

        Everything is orderly and sensible this way. More than just services. Not just one cost trend fits, they all do: houses, colleges, spending per student, income of doctors, whatever I look at!

        How is this not considered significant?

        Why am I the only person analyzing things this way?

        • blumenko says:

          How is that 40% possible, when real hourly wages have only increased 15% since 1996?

          • Charles Engelke says:

            Because NOMINAL wages increased 130% since 1996!

            My entire thesis is that what economists are calling “real” inflation adjustment (based on what a basket of consumer items cost) is not as ‘real’ as scaling by the ratio of nominal wages per hour. The things in the basket are being produced using progressively less work time and occupy a progressively shrinking share of economic activity.

      • Charles Engelke says:

        But it is true that healthcare itself is rising in cost massively.
        That is evident from the increasing share of GDP pie (18% of GDP means the average worker spends 18% of his time working to pay for health care, in 1960 it was 5%). That’s the sign of something that is becoming more expensive in absolute terms–the reverse of agriculture going from 30% to 1% in a century)

        I don’t know where the money winds up, not in doctor’s salaries obviously, but this is not equivalent to the Baumol effect!

        Everything normally attributed to Baumol rises at the rate of what I’d call “false inflation” (the percent at which CPI fails to correctly compensate the real inflation created using monetary policy)

        I have finally found a mention of someone else looking at the problem in terms of time worked to afford things:

        Mark Perry has essentially found that what I would call the the work-absolute price of healthcare in 1958 was only 1/4 what it became in 2012. that agrees decently well with the ratio of 5% GDP to 17% GDP (1960 to 2012) i.e. 1/3.4

        My absolute inflation index using the nominal prices from Mark Perry translate to $4,000 (2018 equiv) per person for health care in 1958 and $10,400 (2018 equiv) in 2012, so 1/2.6
        (Perry gave a slightly high wage for 1958)

        Report commentReport

        Click to Edit – 59 minutes and 18 seconds
        Reply Hide

        • 10240 says:

          That is evident from the increasing share of GDP pie (18% of GDP means the average worker spends 18% of his time working to pay for health care, in 1960 it was 5%).

          The total spending isn’t a measure of the price of a given amount of healthcare. The “unit price” of healthcare would be the total spending divided by the “amount” of healthcare we buy, however that could be defined. We aren’t buying the same amount and quality of healthcare as in 1960.

          • Charles Engelke says:

            I wonder why countries like Singapore have a much smaller fraction of GDP used for healthcare though. They have good results. What are we “buying” that isn’t necessary?

          • 10240 says:

            @Charles Engelke As far as I understand,
            • Americans buy unnecessary doctor’s visits and unnecessarily expensive drugs as a result of low-copay insurance policies. Countries with public healthcare tend to ration these or discourage the• m through long waiting times etc., while other countries with private health insurance tend to discourage them with higher out-of-pocket expenses.
            • Americans tend to get private hospital rooms, which are not standard elsewhere.
            • Even if doctors’ salaries haven’t become higher in America over time, they are higher than in many other countries. Equivalents of the Baumol effect exist over space rather than time.
            • Singapore has a younger population.

    • blumenko says:

      So definitionally in your system average-skilled workers never experience a change in real pay rates. So how can we explore if they are better off today or 50 years ago or in another country?

      • Charles Engelke says:

        Blumenko, you compare how many units of the things getting cheaper they can buy!
        how much less they need to work in a day to meet the necessaries
        how many more things they can buy beside food and fuel (which used to take 90% of average people’s work time, now it’s 15% or so)

        these things differ in other countries and times

      • Charles Engelke says:

        for that matter, how do conventional economists explore if people are better off:
        they look at the same ratio of how much the people can buy at a given time, but they freeze the value of the things getting cheaper in terms of work/time and rescale everything else, claiming that more money has magically appeared for the people to use allowing them to buy more of the things I say are getting cheaper and they say are constant. But that was too confusing.

        Maybe it’s more illustrative to push the extrapolation “ad absurdum” two ways,
        using the conventional, and then my own viewpoints:

        Conventional economics says that GDP per person increases with progress, so eventually, if progress continues indefinitely, the GDP per person will rise beyond Bill Gates’ fortune, and continue increasing toward infinity. The belief that somehow all wages should be coupled to productivity increase means the day will come when the lowest quintile earns hundreds of billions per year. What does that mean for the nature of an economy? What jobs are they working in that pay so much? Do average people actually use any significant portion of their income? I can’t imagine the situation.

        In my model, money only represents the uncompensated work someone has already contributed to another and is now waiting to have redeemed from society at large. There is a maximum potential amount of work per person that can be generated to trade in this way which we can think of as a ‘pie’.
        As labor productivity progressively increases, everyone needs work less and less to get necessary things because fewer people need to be compensated for their work making things one buys.
        There is so much potential human work time left over after a while, new sectors with desirable offerings spring up to induce people to exert themselves again and fill in more of the empty part of the pie. Eventually a rich diversity of desirable devices and entertainments are available, but the empty pie still wins out and leisure grows nevertheless to dominate.

        The infinite end point of my extrapolation is best thought of as an approximation of the Star Trek future: a veritable apocalypse of wage/productivity decoupling: something analogous to the Star Trek Replicator! Virtually an economy without money.
        In Star Trek, Replicators, using transporter materialization technology driven by inexhaustible dilithium crystal power, provide for free, any food, object or known device that any citizen of the federation requests in their own home. Here is the ultimate in “outcome equalization” at last!
        But is there a problem that wages haven’t “kept pace with infinite productivity?” (Actually, wages would still be necessary to employ other people directly or to rent a room or buy land: those will still go for the equivalent hourly rate as today, they have no deep connection to the source of “productivity”)

        Is the GDP of the Federation infinite? Or is it zero? Do you care? How would a billion dollars help you more than free stuff ala “Big Rock Candy Mountain” or Aladdin’s lamp?

        Of course, one can still join Star Fleet to explore, or create art, or provide services to others, but no wolf threatens at the door if you don’t.

        • blumenko says:

          In the standard economic view, yes everyone in Star Trek has infinite income which they spend on basically free replicator stuff, like Lamborghini-equivalent spacecraft, whatever food they want, etc. which if you add it all up in today’s prices costs a hefty sum. If people still needed real estate or to employ other people directly, then GDP wouldn’t be infinite just very large, and a huge share of people’s incomes would be tied up in those things, and hence inflation would also be high.

          • Charles Engelke says:

            Yes, agreed,
            and I think that is a nonsensically unphysical claim to be making.

            So I am against the standard economic view.

            Money is for facilitating the trading of work between people.
            At anytime, I believe, it should be tied only to what value people have given to others and not yet been compensated for in kind.
            It shouldn’t be tied to an abstract valuation of some intrinsic worth in some totality of things.

            We treat water as if it were free (but it was rare and hard to get clean water in the past) If eventually, we do get water for free from taps, does that mean an economist from king Herod’s distant past can legitimately prove we essentially have infinite income?

          • Charles Engelke says:

            I’d rather define money in terms of average compensation for work hours, and evaluate economic progress by using something like the hours needed to produce things in a CPI-like basket. This would shrink with productivity, so you might prefer the inverse. Or take the time an average worker had to work to buy the contents of the basket at the reference time (when basket was decided upon), then, in subsequent years, take the new time it takes to buy the basket, and subtract it from the reference time, maybe divide this difference by the original reference work hrs.

            When that indicator gets near unity, you know that particular basket is no longer useful (so you don’t get the infinite inome from water problem)

          • Charles Engelke says:

            Having “astronomically” high inflation in the Star Trek universe would be a stupid price to pay to keep today’s economic analysis based on price of ‘like’ widgets “standard.”

            What is real and constant in all eras is the people’s experience of time and effort, and what that indicates to us about how much they value what they buy with its earned proxy, money.

            The thing to look at is not our valuation of what they were buying, but how much they valued it.

            For now, just take people’s valuation of time: would you spend an hour, say in jail, in return for a dozen eggs? Then the next hour for a pound of pork chops, the next for a pound of butter,…and so on for every item on your grocery list?,
            At some point you need your time for something else!

            But the average person in 1935 had to work about that long for each of those items, while in 2019 the average person trades only 2 or 3 minutes for each.
            Doesn’t that mean more about the comparative ‘cost’ of eggs then and now than whether they are grade A in our terms?!

            If there is a potato blight in 19th century Ireland, the amount paid for an intact potato cannot be used to value the currency of the time by comparing it with our present potato pricing!

            the goods are not what we value, its our need or desire for them, in context of our time, and that is revealed by how long we work to get them

          • Charles Engelke says:


            If people still needed real estate or to employ other people directly, then GDP wouldn’t be infinite just very large, and a huge share of people’s incomes would be tied up in those things, and hence inflation would also be high.

            But this statement is a description of cost disease and so answers Scott’s original question!

            Clearly there is no market competition going on in the Star Trek limit. Nobody can stop giving massages and be better paid working in a replicator!

            Just a definitional problem with calculating ‘real’ GDP.
            =THE cause of Baumol effect.
            (Taking limits is often clarifying)

          • Charles Engelke says:

            You can say you just prefer the standard economists’ way of describing things, even though its clearly not the only way, it’s the established convention,

            But, then accept the illusions that it creates with open eyes. Don’t run around trying to cure a “cost disease” and make policies in the Star Trek universe to reduce the pay of masseuses progressively to zero thus staying proportional with the stuff that comes out of the replicators.

            The same goes with trying to avoid having income concentrating in a small percent of people when the cause of productivity increase is an efficient network (rather than shrinking manufacturing cost per unit): J K Rowling made several billion dollars due to the “Harry Potter” boom while other children’s book author’s income stagnated! What can be going wrong? Does she have too much political influence?

          • Charles Engelke says:

            But, perhaps I really should make more of a case for the naturalness and morality of keeping average wage constant when adjusting for inflation, or targeting monetary policy.

            I think Frederick Soddy (the 1921 Nobel Prize winner for discovering radioactivity = transmutation of elements with release of ‘atomic energy”) was right, when venturing outside his specialty, he posited that fundamentally, money is a way to allow work specialization through the barter of equal work values within a self-organizing cooperative. As with all cooperation, the key is trust.

            When someone has contributed the fruits of work and skill to another and received only a token IOU of equal value redeemable from society at large, he needs confidence that when he does require something concrete himself, the IOU will still equate to the value he trustingly contributed.
            The value of a kind of work is shown by how long the average person would work to get it, this is settled through market forces, not statisticians.

            When any inflation occurs relative to average work, it is amoral and destabilizing. Yes, I know average people won’t have an incentive to spend their money quickly.

            That is not a problem compared to what happens with the richer people who can collect capital for future investments or emergencies. But their money is made to evaporate by government monetary policy, and their only fully reliable recourse is to buy land which on average retains its intrinsic value relative to work (see figures for average house price in my google doc link).

            That means that rich people buy up more of the best land than they otherwise would want, which raises prices and reduces options for the less well off. It also leads to speculative bubbles in the most desirable places where demand for a safe place to store money increases demand and the rising price encourages more buying since the value is actually increasing relative to average wage inflation.

            But this is not real value, it is psychological future value.
            Humans bind future and past in their planning, but future is not known definitely. The positive feed-back in all evolutionary systems is vital: “this worked — so do more of the same –repeat”, that is what drives real productivity innovations and experience curve effect, and new effective structures for production are invested in quite properly by people with excess capital as their superior future potential shows through success.

            But pseudo-success trends, fueled by Ponzi-like structures can induce the same automatic response, even when there is no real source of value increasing.

            And nothing can drive such impressions of pseudo-success like inflation induced distortions along with its smothering all sight of the real trends

    • Bamboozle says:

      Wish this had more responses would be interested to see what people think of this.

  28. adaffodil says:

    This issue has to be considered at the margin. The marginal increase in salary for the marginal worker is the source of rising costs, not the average salary. An employee is pulled into the sector when the wage exceeds the wage in a comparable sector after subtracting whatever transaction costs exist in changing sectors. The wage difference could be quite small or large for the marginal worker depending on the transaction costs involved. The total difference for society is the share of workers employed in those sectors times the wages in those sectors. If the wage increase isn’t in doctors salaries but we pull in lots of additional nurses that have lower transaction costs to change sectors, then the average wage doesn’t need to rise much or at all. It could even fall as the people drawn in earn less than average but more than they would elsewhere.

    If we look at the total labor resources dedicated to these sectors, those in fact have risen dramatically. Over the last 50 years or so, the share of total US employment in education and health services has risen from roughly 6% to 16%. This is the baumol effect in action. Low productivity sectors take much more labor input to increase output than high productivity sectors. Measured average wages within the sector have almost no explanatory power.

    • baconbits9 says:

      This is not the Baumol effect, the Baumol effect does not claim that we are going to demand more violin concerts, only that those we do demand will cost more.

  29. warrel says:

    I’m not sure if ‘rising wages’ here means rising wages vs. productivity, but I DO remember Tyler Cowen saying that he thinks the problem is on the Productivity side: He thinks the rises in productivity aren’t Real.

  30. danjelski says:

    So I’m really confused. The Baumol effect says that while Widget productivity over time goes from 5 widgets/hour to 500 widgets/hour, string quartet productivity doesn’t change at all. To my mind, this means that widgets are “high productivity” while quartets are “low productivity.”

    The analog to string quartets are professors–as one of my colleagues once joked: “I’m going to increase productivity this semester by talking 15% faster.” Joke notwithstanding, a 50 minute lecture still takes 50 minutes, even after all these years. College professors are low productivity (compared to widgets), which is why they might be subject to the Baumol effect.

    Yet throughout his piece, Mr. Alexander refers to “education” and “healthcare” as “high productivity.” Yet they (or at least education) are precisely not–at least the way I understand it. So I found the piece very confusing–I don’t understand what he means.

    Please explain.

    • JPNunez says:

      I think there is a coma, but it indicates that he is listing stuff so

      My counter is that wages in high productivity sectors, education, and medicine are not actually rising.

      Means “high productivity sectors, AND ALSO education and medicine”, implying those are not high productivity.

      If you check the previous article on this

      Put another way, a violinist can always choose to stop playing violin, retrain for a while, and work in a factory instead. Maybe in 1826, when factory owners were earning $1.14/hour and violinists were earning $5/hour, so no violinists would quit and retrain. But by 2010, factory workers were earning $26.44/hour, so if violinists were still only earning $5 they might all quit and retrain. So in 2010, there would be a strong pressure to increase violinists’ wage to at least $26.44 (probably more, since few people have the skills to be violinists). So violinists must be paid 5x more for the same work, which will look like concerts becoming more expensive.

      This should happen in every industry where increasing technology does not increase productivity. Education and health care both qualify. Although we can imagine innovative online education models, in practice one teacher teaches about twenty to thirty kids per year regardless of our technology level. And although we can imagine innovative AI health care, in practice one doctor can only treat ten or twenty patients per day. Tabarrok and Helland say this is exactly what is happening. They point to a few lines of evidence.

      So yeah, education and medicine are not high productivity, the opening to this article is just a little confusing.

  31. JohnBuridan says:

    Talking to Mrs. JohnBuridan this morning she pointed out some increased efficiency in teaching over the past 75 years. Thanks to calculators more students can reach Algebra 2 and beyond, so now Algebra 2 is an expectation for all students: even the GED equivalent tests have some Algebra 2 on them. The entire system has been trying to move complex math concepts to lower and lower grades for years and has been astonishingly successful – at least for some students (data forthcoming?).
    Schools have become much more efficient at working with developmentally challenged students (perhaps), which for decades was an ignored issue – some students need a lot of additional help. Various student services are now offered…

    On a different note, I wonder how government subsidy rates changing over time affected education. AFAIK, before Reagan’s tax cuts gifted education in America was large, then it shrivelled up. But expenditure on the left tail stayed the same or grew. Ahg! I see why Scott is obsessed with this problem. It feels like the puzzle pieces are there but we can’t figure out how it all goes together!

    • The Big Red Scary says:

      I used a calculator in algebra class, but I don’t remember it being good for anything besides 80085. I do however remember y=mx +b and (-b \pm sqrt(b^2-4ac))/2a. Do they teach those in algebra anymore?

      Could you ask Mrs JohnBuridan to clarify?

      • JohnBuridan says:

        I sought clarification: calculators have helped more students get into higher level math than they would otherwise be able to get to because mastery of math facts can take many more years for some students, especially ADD and dyslexic students. This makes lots of problem solving – especially problem solving in a timely manner difficult. Throughout high school there is a lot of math that is easy for students who understand the integers to intuit and solve quickly, but slower students are not be able to solve those same problems in a timely manner and so never learn to enjoy the beauty of math.

        I am reminded of this story about a student who randomly did great with math when she didn’t have to work with integers and this paper about numberless word problems.

        The formulas you mentioned are of course necessary to solve problems and students still learn them, but hopefully they learn why they are true, and so by understanding the concept know when to implement the procedure. The AoPS curriculum is motivated by this thinking: calculators can calculate, students can problem solve. Consider this humorous story in which Richards gets mocked for having lots of formulas at a Math competition.

        • The Big Red Scary says:

          Thanks. That didn’t occur to me. I guess I’m an old geezer, because my algebra teacher kept the calculators in his desk and only took them out for special activities.

          It’s an empirical question whether there are
          students who are more or less incapable of
          fluency in integer arithmetic, but otherwise can do math. If there are, then I can imagine letting them use the calculator could be helpful.

          In his reminiscences about Stanislaw Ulam, Gian-Carlo Rota writes about Ulam’s extreme laziness, which seems to have been an in-born trait, as well as his relatively short attention span, which came about after having his encephalitic brain sprayed with anti-biotics. This gave Ulam rather peculiar intellectual qualities that allowed him to fill an important niche. I suppose Ulam would have been happy to use a calculator.

          I once worked as a tutor in a program “JUMP” that took the opposite assumption to that of the “problem solvers”: many difficulties with mathematics are psychological, “hard'” arithmetic can be learned in tiny steps, and that doing so can boost confidence and remove psychological obstacles. Here’s the link to

          Of course, everyone agrees that at the end of the day, manipulation of symbols without understanding is useless, except for some signalling purposes. The question is how best to achieve understanding. Did one side win the “Math wars”?

          Ultimately, I suspect part of the problem is the focus on how to teach rather than on how to learn.

  32. Tadas says:

    Suppose we live in a world with no inflation (or it’s adjusted for, as most of the analyses assume). We have a worker bee and an artist butterfly. Both make 20 dollars per month and buy 10 dollars worth of art and 10 dollars worth of goods.
    The productivity of the worker bee skyrockets and she can make 10 widgets instead of 1 in the same amount of time. Due to competition between manufacturers the price of a widget drops to 1 dollar. Should the wages of a bee increase in this world? Given a robust competition, no. The profit margin of the factory stays the same, the price of the output drops by a factor of 10, the bee makes the same 20 dollars a month.
    But now the insects consume 10 dollars worth of art and 1 dollar worth of goods, all things being equal. Baumol disease strikes again, except it doesn’t really. Relative price of art increased, compared to the goods, but the absolute art price didn’t have to go up in this toy example.
    The salary of the more productive worker also doesn’t have to increase with the productivity. Why? Well, because in dollar terms the worker did not become more productive. That gives an insight of how the productivity can be (and is) easily mismeasured in the real word – it has to be decomposed into the price inflation component and the actual extra dollar output that was created by the bee. Arguably the extra dollar output is imaginary in most cases, simply because the competition drives the output price to the level which keeps the owner’s profit margins and the worker salaries constant.
    The main reason none of these charts hang together is because they are adjusting for inflation, as if one price or wage inflation rate exists to describe all sectors of the economy. Due to friction in the economic system, you can’t instantly compete away elevated salaries, profit margins, or cost of goods. Some anomalies persist, and they are sector-specific. Yet we assume that we can average apples, oranges, and grapes, and get a meaningful result of inflation which applies to every economic actor, entity, and sector.
    When studying complex phenomena, I learned to slice and dice, decompose into pieces and not compare the incomparable, pull on the various ends of strings tangled into a ball, but never to yield to the temptation to produce a single number. Economics is not the same as finance. Implied volatility of an option is very useful when comparing different securities with different prices and expiration dates, but the real economic system is not a basket of options.

  33. bernie638 says:

    An interesting data point since you highlighted physician pay 1973 to today, is military enlisted pay. The numbers are easy to find and coincidentally 1973 was the year the draft ended. Looking at your physician pay example, today they make 5.64x what they made in 1973. An E4 with 10 years of service made $5,556 a year in 1973 and $31,967 today for an increase of 5.75x. An E5 over 10 was $6336 then and $40,510.32 today for a 6.39x increase. An E6 over 10 was $6693.6 then to $43,874.76 today for a 6.55x increase.

    It doesn’t look like physician pay kept pace with the pay raises for military enlisted people.

    • Steve Sailer says:

      Good point.

      The military are not poorly paid anymore.

      On the other hand, they tend to do more tours in war zones than in the past.

      • bernie638 says:

        And the benefits are the best! Free housing, free food, no copays or deductibles for healthcare that includes all the motrin and foot powder you can ask for.

        Still, framing matters, they aren’t poorly paid on an annual basis compared to median wages, but hourly wages are still below the legal minimum much of the time.

        I just thought it was a good comparison for people who think that doctor pay is the reason health care is expensive.

        Also I classify the military as the ultimate service industry (it’s right there in the name), and the destructivity of the mark one model zero grunt hasn’t changed since the 70s (they use the same rifles even).

        • Matt M says:

          but hourly wages are still below the legal minimum much of the time.

          Eh, only for the really junior enlisted, and even then only, as you say, when you conspicuously don’t include all of the benefits (many of which go well above and beyond the standard sort of “benefits” one is used to thinking of in the private sector)

  34. Swami says:

    It is partially about how you calculate inflation, which explains around 35% of the problem depending on who you believe. But we’re comparing wages to the cost of education/medicine. As long as you’re using the same deflator for both of them, you’re fine.

    I am not an economist, but this does not strike me as being correct. The difference in inflation rate occurs when you switch from CPI to PCE. And PCE reflects that people adjust to variations in prices by changing their choices of what to buy. Thus the price of medicine can go up by 50% using the CPI, but the salary involves a human making trade offs and thus would use the PCE. So, as I understand it, we would not use the same inflator for them both, and arguably should not be.

    A real economist should feel free to correct me or clarify, though….

  35. jgr79 says:

    I think you’re wrong about your interpretation for 2.

    Let’s say I spent $100 on education in 1990. Since salaries plus benefits were ~82%, that means I spent $82 on salaries.

    Now let’s say my education budget has doubled to $200 (constant dollars). Since salaries are ~80%, I’m now spending $160 on salaries.

    Of the extra $100 I’m spending now, $78 are going to salaries. So in my example 78% of the increase in price was due to rising wages.

    Ie more than 3/4 of the increase in price is due to wage increases. If we’d held wages constant, we might expect the price increase to be about 1/4 of what it really was.

    This seems to confirm their claim – the majority of the price increase in education is due to wage increases.

  36. Charlie Lima says:

    I have actually have a hard time buying that physician productivity has not increased.

    For one thing, doctors simply do more today. Back in 1960, we had basically no treatments for heart attacks, few medications to prevent them, and not a whole lot to mitigate symptoms afterwards. Today we have the miracles of the cath lab. We can use thrombolytics. We can do coronary artery bypass grafts. We can drop in an LVAD if things are dire enough and possibly even implant a second heart in series to the first. All of these interventions require specialists who simply did not exist in 1960. So yeah we have more cardiologists per capita, but Americans now rarely die in their 50s from heart attacks.

    Nor is this unique to heart attacks. Whole specialized fields of medicine have arisen since 1960. Emergency, critical care, interventional radiology, and many sub-specialties basically did not really exist in 1960. So again you need more physicians to provide additional services.

    Then on the flip side “healthcare” is one of those terrible to metricate concepts. What do we really want from the doctor; we want to become healthier than if we did not spend resources on medical care. What we, properly, should be measuring is for a given a patient how much mortality and morbidity have our physicians prevented?

    I would submit that our whole healthcare system is preventing massively more mortality and morbidity than in 1960. In 1960 ~10% of the adult population was obese or worse; today about ~40% is bad. In 1960 the average American married in their early 20s and stayed married. This is associated with vastly better health outcomes and cheaper healthcare (having a spouse who can change bandages can save you a very expensive stay at a skilled nursing facility for instance); today average age of first marriage is approaching 30 and record numbers of people are single throughout life. In terms of mental health, Americans’ baselines have never been worse. We have ever increasing rates of depression, anxiety, and everything else. Maybe these are just changes in diagnosis standards, but somehow we have a very rapidly rising suicide rate. Mental health is an independent risk factor for basically causes of death. We have moved the needle on smoking but otherwise the bell curve for social determinants of health has shifted in the wrong direction.

    A physician from 1960 seeing today’s average patients would wonder why he is treating all the hard cases. I expect that he would quickly alienate well over half of them (who would stop taking medications or listening to his advice) and we see a steady rise in morbidity and mortality for his patients. If you put a doc from today back in 1960, she would save countless lives with small practice changes (let alone introducing new technology) and be amazed that all her patients are basically healthy.

    Physicians (and nurses and all the rest) simply accomplish more today. If nothing else, everyone who does not die of the things that killed people in their 50s and 60s back then becomes much more resource intensive to treat. Bypass surgery, for instance, has created a lot more people surviving their heart attacks who need various degrees of cardiac follow-up. Dead patients just need a quick visit to the pathologist.

    The other major thing that makes me wonder is how much the cost of becoming a physician has changed. In 1960, the median medical school charged ~$1050 per year at the expensive private schools. Using CPI adjustment we are looking at around ~$9200 per year in 2019 dollars. Median public medical school tuition is now north of $32,000 per year. Doing a true apples to apples for a private school is north of $50,000. Ignoring increases in housing costs, we are looking at the capital input required to become a physician basically quintupling.

    Except, of course, that things are worse than this. Back in 1960, many states allowed one to begin practicing medicine after a single intern year and then start making a reasonably high percentage of the average physician’s income. Today, you need three years of residency even for the most basic family med or internist options. This means that physicians today not only have massively more debt, but that it takes them longer to pay it down. We measure the inputs, we should expect physicians to be much more expensive as they tie up much more time and capital in their training. And lest we forget, it is becoming vastly more common for medical students to need a year or two of research or work experience before medical school and another year or two after residency for fellowships. Once you include changes in retirement patterns, you end up with sizeably shortened career span.

    Paying off a quintupled price tag with fewer years of earning seems like a pretty straightforward way to raise prices.

    Seems odd that we are talking about price while ignoring both quality of services and the capital inputs required to train the service providers.

    • Steve Sailer says:

      “I have actually have a hard time buying that physician productivity has not increased.”

      I’m impressed with the steady increase in quantity and quality of minor medical techniques I witness in periodic doctor’s appointments or, especially, visits to the emergency room.

    • Steve Sailer says:

      My dentist is about the same age as me (60) and grew up in the same neighborhood in a then middle class and now upper middle class part of the San Fernando Valley. I figure it cost his parents about $6k total out of pocket for his education all the way through UCLA dental school.

      Being Catholic, I went more private school until B-School. My parents probably paid in tuition a total of $3k for 8 years of parochial elementary school, $3k for 4 years of Notre Dame HS, $12k for 4 years of Rice U. (Rice charged a low tuition due to its prosperous endowment), and $3k for 2 years of UCLA MBA school.

      Back then Jewish people like my dentist eschewed private schools as sort of un-American. The turning point was the imposition of racial school busing in 1978. Local Jewish politicians like Alan Robbins and Bobbi Fiedler led the fight against busing, but when they were defeated, Jews white flighted from the LA public school system. These days, the San Fernando Valley is full of very low profile Jewish private schools, although the very public struggle of 4 decades ago has been memoryholed.

  37. Ijime says:

    So, I don’t know if someone has posted this before, I haven’t read all the comments.

    But doesn’t the Baumol effect work just as well on the “investor” side as the wage side? I.E productivity in other sectors rise, increasing the relative rate of profit of those industries. In order for anyone to ever invest in education or healthcare, they would have to increase their profit margins to match. But increasing their profit margins without increasing productivity means raising the price (/lowering the quality or wages).

    This would explain how a Baumol-like effect can happen even if wages don’t increase, although presumably this would cause some other measurable change, and would probably already have been noticed by now.

  38. Worley says:

    I’d recommend using other examples than medical care (in the US) and college (in the US). Those two industries have been the most distorted away from being simple markets by various features. Some of those features are explicitly designed to decouple the choice to consume from the price of the resources that will be expended. Another is that the measure of the product is constantly changing. We demand medical treatments as a right that simply didn’t exist when I was a boy. Similarly, my alma mater (which is an excellent college that in theory shouldn’t have any marketing problems) finds it necessary to ramp up its student amenities and add many more specialties to its academic program in order to remain competitive.

    A better choice for studying Baumol, I think, is public primary and secondary education in the US since maybe 1950. The product that is delivered hasn’t changed greatly and the business structure of the industry basically hasn’t changed at all.

    There’s also the complexity that the US hasn’t been a closed economic system for decades. Looking at wage statistics for the US is like looking at wage statistics for California — you don’t expect them to be the only, or perhaps even the major, driver of prices or consumption for any good or service that is tradable. Better would be to take everybody in the globalized countries (especially including China), and comparing them now with their parents in 1990 or their grandparents in 1960. The media US wage hasn’t changed much, but the median wage of the larger population has increased a lot. The elites of other countries are making a lot more — and they now form a significant part of the demand for US colleges. Conversely, manufacturing hasn’t so much become more productive as simply replaced an expensive set of workers with a cheaper one.

    Instead of focusing on e.g. manufacturing workers, who used to be elite with respect to the world and now are not, it might be interesting to look at the wages of Ph.D.s or M.D.s or some other occupation which hasn’t been much subject to foreign competition. If you use their wages as a baseline, do the prices in different industries show the Baulmol pattern?

    • Worley says:

      I have this somewhat silly urge to say that looking at average pay rates in the US over the past 50 years is like lookat at average pay rates in Flint, Michgan over the past 50 years, and saying “Have string quartets gotten more or less expensive?” Flint used to have a lot of well-paid unionized GM workers. Then it didn’t. String quartets got a lot more expensive relative to Flint wages. But since all of that auto work went elsewhere in the same economy, Flint is anti-representative of the entire economy.

  39. Steve Sailer says:

    I’d like to see somebody collect a lot of examples of trends in compensation for jobs that are relatively homogeneous over the years, such as starting salaries for associate lawyers at Cravath Swaine, starting salary for a Dallas public school teacher, minimum salary for a Major League Baseball player, and minimum salary for a Chicago Symphony Orchestra musician. And then put them in rank order of change over time. Would the rank order fit what is predicted by the Baumol Effect? Or is something else going on?

    I often find that putting things in rank order quickly generates better hypotheses about what are the driving factors than does theorizing from abstract models.

  40. Urstoff says:

    What happened to the explanation I heard a lot around the time of the ACA that healthcare costs have increased simply because we consume more and use more expensive medical technology? Cost per service has increased because input costs have gone up (more expensive machines), and total number of treatments have gone up because our mongrel system doesn’t really have an way to effectively ration healthcare. Is this hypothesis one that didn’t stand up to scrutiny?

  41. zstewart333 says:

    Also, the Baumol effect only works if the market sets your salary in the first place! Right now the supply of doctors is limited by licensing issues and bottlenecks in the medical education process; that keeps salaries high. Why would the Baumol effect drive that even higher? If doctors’ salaries didn’t increase in keeping with the highest-productivity industries, would medical schools sit empty?

    The Baumol effect has TWO parts as T&H emphasize repeatedly in the book. Part 1 is that the relative price between two sectors increases, Part 2 is that we can still get more of everything, just with higher tradeoff costs.

    As active agents, Medical Schools can set the level of undersupply to keep doctor salaries at a fixed ratio above the free market rate. As Baumol effect Part 2 causes society to demand more resources on doctors, medical schools can provide more doctors without providing enough doctors.

    So we get more heathcare in absolute terms, doctors make more than they would in a free market, and we feel that doctors are very expensive when priced in cars and hamburgers.

    US physicians per capita rising over time:

  42. ronpdavis says:

    Couple thought.

    1. Glad you got point #3. This is, I think, too unappreciated when it comes to cost disease. Even though more absolute buying power remains in hand after paying for things like education and healthcare, that is only true across society and not for a given individual. I think the literature has failed to grapple with the meaning of this. If healthcare and education become 70% of GDP and the median pay is, say, 70% of the average productivity per worker, then half of society will only be able to pay for education and healthcare.

    The social implications are enormous…as that happens, it is unlikely that a larger and larger slice of society will be able to afford these critical services, and that means taxes as a percentage of GDP will go up, much much higher (or people will be starved of essential services). There will be enough surplus in absolute terms, but will taxes in relative terms cause serious class conflict?

    2. I haven’t read this book, but at a quick glance I think the analysis is too wage focused. Real wages tracking with productivity growth doesn’t seem to be the linchpin of Baumol’s argument. The long term effect of productivity increases is falling prices for the outputs of that industry. If demand for less fast growing industries is inelastic, this means their prices will rise relative to the slower sectors. Whether wages rise will be more complicated: does the increased productivity benefit holders of capital or the workers (depends on politics, and what drives the increased productivity). Does consumer demand rise as fast as the prices fall? If not, employment I’m that industry will fall/wages will decline. (If the price falls by half and demand only increases by 10%, employment falls).

    I suspect wages are flat and service prices are relatively high because productivity per worker has increased while demand for eg heathcare has been stable (or growing!), which would create the cost disease effect. But politics has decreased labor’s bargaining power, the role that technology (and that capital) has played in productivity increases is high, and the real benefit to workers has been in fast deflating prices in areas where productivity growth is high (e.g., computer chips). Interestingly, these tend to be the most difficult areas to think about when we look at inflation.

  43. Charles Engelke says:

    Any physicist who has worked in a University has received their share of ‘crackpot’ missives explaining someone’s new unified theory, or why Einstein’s relativity theory is provably wrong. Some are by obviously intelligent people who just don’t know the huge background body of math and observations and how tightly things have to hold together. Usually their ideas seem more metaphysical than physical, not something one can tie down for testing.

    Now, apparently I, with my repeated insistence above on using work/hr equivalent inflation, am in he crackpot shoes vis economics. But, I feel I have explored the logical consequences of my assumptions, as well as those of the conventional economic inflation definitions, and mine seemed to be the ones giving sensible ‘physical’ answers while theirs seemed supportable only through rote metaphysical definitions. “Hedonic adjustment?”, really?

    For all its high mathematics, economics, like most social sciences, has never gone through the “predict-test-falsify” positive feed-back driven evolution and firming up the physical sciences have. So, it’s about time to start.

    I propose the following physical test as a challenge:

    The BLS CPI says that a dollar in 1935 was worth the equivalent of about $18.23 in “real” 2019 dollars

    My “constant value for average work/hr” index says that it is worth about $92 “real” 2019 dollars.

    To find out whose conversion better deserves the moniker of “REAL” we should set up a REAL-ality situation such as in old reality shows of the “1800 house” or “pioneer house” ilk.
    Call it “1930’s Farm”

    Give a team of economists a typical 1930 farmer’s cash reserves (and monthly payments on loans) scaled by CPI to modern $2019 money, plus a typical allotment of land and equipment. They then have to use the authentic 1930’s methods on pre-rural-electrification style farms which required 20 times as many farm workers per consumer served, one finicky old gasoline hungry tractor per 40 farm workers, horses, primitive plows, ineffective insecticides such as powdered lead of arsenic or sprayed kerosene, pre “green revolution” fertilizers, etc., etc. … in order to grow enough food to feed and pay the farm workers and cover all costs. They have to transport their produce to markets in small trucks lacking refrigeration (with maybe some packing ice) absorbing all spoilage costs.

    The question is whether they can break-even or make any profit selling to 2019 customers at the official CPI version of the converted 1930 prices, and if not, what price would!

    We can guarantee the purchase of what food gets to market unspoiled at any price, the question is to find what level reaches break-even for the farm in real 2019 dollars in 2019. The CPI scales the 1935 food in city markets to around today’s Whole Foods prices (though bacon is cheaper than at any 2019 normal average supermarket (guess they made a lot of it in 1935, having no refrigeration). CPI also claims farm workers got $7,000 a year in real 2019 equivalent. Note, the prices for food make sense for our experience of actual 2019, but not the farm worker’s salaries.

    If the test farm can not break even trying to sell at what economists say were the actual 1935 prices in 2019 equivalent (as I predict), then in what sense can they label those 2019? I predict that the break-even prices would need to be $30 for a dozen eggs, $25 for a pound of pork chops, etc. Shockingly high for today’s supermarkets (but the real experience of 1930’s shoppers, and the reason people looked so gaunt in the old photos)

    • The Nybbler says:

      This is real compensation per hour for the “business sector” — the entire economy less government, nonprofits, and private households. That is, it’s average compensation per hour deflated by the conventional index of inflation. Note that it’s going up. I gather that you used “Personal Consumption Expenditure: Services” to create your “constant value for average work/hr” deflator; this is simply wrong, as consumer spending on services is not at all the same thing as wages.

      Also, using a farm is probably the absolute worst way to do things. The usual CPI used is the consumer price index for all urban consumers, and many measures of the economy either exclude or treat farms differently.

  44. Charles Engelke says:


    My point is that using CPI adjustment is wrong, it’s an arbitrary convention and it distorts what is going on physically in the real world. I am saying the ‘real wage’ plot you reference is not ‘real’ it is made-up and refers to nothing concrete. But, yes, the fault in my eyes is with professional economists, not you.

    The original selection of a CPI basket of goods generated using evolving technologies requiring progressively less work (& thus consequently easily glutting and saturating their market) ensures that the thing you are scaling by is a shrinking portion of the economy, thus the wages have to seem to be rising.

    If the price of electricity, for instance, had been used instead as the dominant component of the CPI basket, the ‘real’ wage plot you linked would have been level like mine. But if “Mr Fusion” from Back to the Future ever becomes a reality, electricity would become as bad standard of comparison across time as food and manufactured goods are.

    What is real and constant in all eras is people’s experience of time and effort and what that indicates to us about how much they value what they buy with its earned proxy, money.

    The thing to look at is not our valuation of what they were buying, but how much they valued it.

    Forget the effort for now, just take people’s valuation of time: would you spend an hour, say in jail, in return for a dozen eggs? Then the next hour for a pound of pork chops, the next for a pound of butter,…and so on for every item on your grocery list?,
    At some point you need your time for something else!

    But the average person in 1935 had to work about that long for each of those items, while in 2019 the average person trades only 2 or 3 minutes for each. Can’t you see that makes a bigger difference than the hedonic quality of the product as assessed by economic statisticians?

    If there is a potato blight in 19th century Ireland, the amount paid for an intact potato cannot be used to value the currency of the time by comparing it with our present potato pricing!

    On a future moon colony people would need a large fraction of their money earned just to afford water and air, for which we pay almost nothing. Does that mean moon colonist’s work is worth nearly zero and they have no GDP?

    Time is the substance of life, stop devaluing past generations’ lives.

    on what I used for calculations:
    An “Absolute Inflation” index based on comparing “pay per hour worked of average worker” can be obtained to a decent approximation in many ways (even just using GDP/person is infinitely more informative than expert selected CPI baskets with hedonic adjustments), For instance, I’ve done it starting from average nominal wage plus compensation

    (but it’s a little harder to find historical records of the benefit fraction than to divide ‘services’ fraction of GDP by the ‘services’ fraction of employment)

    But as far as I know we have to calculate this stuff ourselves one way or another starting from each year’s nominal GDP, current population, fraction of population in labor pool, fraction of these presently unemployed, hours worked, etc.

    find calculate it ourselves from nominal statistics for GDP, fractional employment, percent unemployment…

  45. yaolilylu says:

    I feel like changing standards of practice does not get mentioned enough in healthcare. I’m a chemotherapy nurse, 40 years ago our nurses used to smoke as they mixed chemos themselves without gloves on with splatter everywhere, it was fast, efficient and unsafe, and now we have specialized technicians wearing expensive PPE who are not allowed to mix chemo at all if the ventilation hood was even slightly below regulation airflow speed, so it’s amazingly safe and we can’t get drugs on about 20% of days, resulting in major inefficiency. Years ago a TPN bag that got punctured would get a band-aid taped over the hole, and ran on gravity. Now the slightest hint that the central-line might not be working perfectly and you need CT scans and expensive de-clogging medications before you can use it at all. I believe that these changes in practice results in only modest improvements in safety because the staff are good at adjusting behaviour to compensate for risk, but now the system is built with assumption that a) any staff might be an idiot and b) no amount of risk is acceptable. We are throwing vasts amounts of money to prevent problems that are very serious but also very rare. Overall the effect is greatly increased cost with moderate benefit to patients.