That graph is fairly deceptive. While labor's share of national income has gone down slightly (mostly post 2000) it's not nearly as dramatic as presented there.
I think one of the bigger sources of the discrepancy is comparing mean productivity to median wages. If productivity and wages were both normally distributed, this would be fine. But they are not.
This article presents the reasons behind cost disease as if his interpretation is accepted fact, but that's not the case, right? Elsewhere, I've seen a lot more disagreement among smart people.
For example, he says we have to pay band members a lot more now or they'd leave for better jobs and that this also explains the rise in education costs over the last few decades. Education costs have risen about 2.5x over the last 40 years, after adjusting for inflation. Has teacher pay risen 2.5x after adjusting for inflation?
Baumol's argument is that services will become relatively more expensive than goods, because productivity increases come faster in goods than services.
This is obviously true, so yes, the article presents it as an accepted fact because it is. We have hard data for labour productivity by sector, manufacturing output, manufacturing employment, etc. We know the number of hours that goes into making a car versus giving a haircut, and how those numbers have changed over the past 40 years.
> Education costs have risen about 2.5x over the last 40 years, after adjusting for inflation. Has teacher pay risen 2.5x after adjusting for inflation?
No, which makes it clear that Baumol's cost disease is not the only driver behind education costs. :)
Quick question for anyone who actually studied Baumol's theories:
Did he bring into consideration how some services, such as education and health care, exist in an awkward limbo between services we pay for (right/libertarian view) and services we're entitled to (left view), resulting in a smash up of public subsidies and political quicksand?
At face value, his theory is totally reasonable. BUT it seems totally reductionist to claim his drivers of cost are the only drivers (even calling them primary drivers seems foolish to me).
The article points out that there are many non-governmental professions that show this effect: musicians (look at the prices for concert tickets), post-secondary education, which is not usually free, though it may be subsidized, restaurants...
In many places, loans for university education are guaranteed by the government. So universities have no worries for charging absurd prices for useless degrees. Even if the graduates go bankrupt the university still gets their money from the government. (And in some jurisdictions bankruptcy doesn't discharge student loans.)
The GP raises a legitimate point that government involvement in just about everything muddies the analysis.
I'm having a hard time understanding the core of Baumol's theory as Vox is laying it out. I identify a few premises:
1) Rising opportunity cost in the labor market raises the boats of all jobs (within a given grouping of jobs with low friction of employee movement)
2) This rising opportunity cost occurred (in the U.S.) due to manufacturing, which was well paying.
3) Consumer goods prices are more affordable due to increased capital efficiency relative to income.
4) The extra discretionary dollars from (3) are being spent on labor intensive goods, which cycles back to (1) again.
One point on which this loses me is the apparently hidden assumption that involuntary un(der)employment is low and stably so, and what markers of inflation we're using. I think there's an unexplained paradox: if we're losing goods-producing jobs, why isn't the competition in the service sector a countervailing force driving down wages?
Not trying to be a critic, I just want to understand this better.
First off, let's consider an economy with a single good (widgets), and no services. Any attempt to measure the economy as a whole (eg, to calculate GDP) is simply going to end up being the number of widgets produced per year.
Further, the total compensation for all factors of production is going to be equal to that same number, which in turn means that any employee's compensation (whatever unit of currency is used) translates into a share of the widgets being produced.
If everyone gets more efficient at producing widgets, the economy is wealthier (and GDP goes up). And each employee is (all else equal) receiving more widgets as their share. Conversely, if we assume that nominal wages have stayed constant, the price of widgets has gone down.
So that's a basic framework we have: Productivity gains = a fall in real prices = a growing economy = an increase in employee compensation. These are all equivalent things; different sides of the same coin. If we're getting better at making widgets, then widgets will be cheaper, which means we can afford more widgets for the same amount of (real) money. That's what economic growth is.
Second, let's consider an economy with two goods: Widgets and sprockets (and again, no services). Let's say half the work force is busy making widgets, and the other half is busy making sprockets. If we get better at making both, but we become much better at making widgets, then the price of widgets will start to collapse. Where 1 widget used to barter for 1 sprocket, soon it'll take 10 widgets to barter for 1 sprocket. Which is another way of saying that the price of sprockets will rise steeply.
That's Baumol's Cost Disease. If you have multiple items, and they experience different productivity growth, the ones which experiences relatively lower productivity growth will become relatively more expensive.
In addition, if you want to take the widget and sprocket example a step further, you might wonder what happens if productivity gains mean we're making more widgets than we strictly need. We'd probably see layoffs, as productivity gains mean fewer and fewer people are required to produce the widgets the economy needs. And of course the people left would probably earn very good wages, since they get to share the returns from the entire widget industry. Although given the high price of sprockets, they might not feel that rich. But where would the laid off people go? Well, into the sprocket industry; it's lower productivity means it can soak up more manpower. But with the returns from the sprocket industry split across more people, wages would be lower. Luckily, those wages would go further when purchasing cheap widgets, but high-price sprockets would be harder to obtain.
Which is, if you squint a bit, basically what we see, with widgets taking the place of the manufacturing sector, and sprockets the services secotr. US manufacturing output is high and (the financial crisis aside) keeps getting higher; US manufacturing employment is low and keeps getting lower. Manufactured goods are cheap and getting cheaper; services are expensive and getting more expensive. Skilled engineers command excellent wages, but struggle to find material goods to spend it on. Workers in the service sector suffer from low wages, although their material standard of living is still good. Etc., etc. Sound familiar?
Productivity gains are always a net good in total, however distributed. But the skewing effect of unequal productivity gains can be extremely wrenching. Telling someone that, actually, the machine shop they worked in is closing, but they can retrain as a hairdresser doesn't make people happy. Nor does pointing out that, even on the wages a hairdresser makes, they can afford a larger flat screen TV than the one they bought three years ago. They don't want a bigger flat screen TV; the last one is actually fine. They don't want to be a hairdresser with a giant flat screen TV; they want to be a machinist. With affordable health care. But that would require figuring out how to make sprockets as efficiently as we've figured out how to make widgets, and we haven't done that yet. (Although the US does seem uniquely bad at it, so maybe we can at least try to be less bad...)
Thank you. This helps me better articulate the thorn still stuck in my side.
Putting together paragraph 2 and paragraph 4 from your response, why don't the following things happen:
The widget company owner starts laying off widget makers to increase profit he can get the same output. Is there no diminishing margin utility for buying more widgets? Paragraph 2 by itself implies that the cost of widgets falls, if wages aren't rising and these increased widgets are being purchased. There's got to be some movement in one or both of cost and salary metrics in order to increase widget consumption if the employer is going to ramp up production so long as the net cost of each widget is positive. What's the incentive otherwise to greater production even if the capacity is there?
Paragraph 4 adds another good with these same questions I didn't feel the article answered well, and introduces the variable of product exchange rates. What unsettles me here is the obfuscation between the total resources poured into the industry and the per-person expenditures, especially in service-oriented work where the productivity isn't simply "widgets" or "sprockets".
Teaching is a great example: the number of teachers (in the US) increased about 48% from 1988 - 2012 but the population increased 29%. Did we get voluntarily make education "less productive" by increasing the decreasing the student:teacher ratio? Did we get better educational quality for it? Not so easy to say.
Could it not be that musicians were paid more by the mid-20th century because there a longstanding oversupply had been steadily declining due to cultural factors, or that concert financiers realized people valued and would pay more for music than had been assumed, i.e. basic supply and demand equilibration?
I'm trying not to be ornery or argumentative, I'm just having a hard time understanding what Baumol's assumptions were on these points, and I think they're important.
> The widget company owner starts laying off widget makers to increase profit he can get the same output. Is there no diminishing margin utility for buying more widgets?
In the simple example where widgets are the entire economy, this doesn't work, because profit is defined in terms of widgets, and because we can assume that there's no diminishing marginal utility for economic production as a whole. At the level of the entire economy, you never become richer by cutting production. (Because again, if the only good or service in the economy is a widget, then you have no way of expressing your profit or wage except in the number of widgets you have. If total production declines, the only way for you to increase your compensation is by obtaining a larger share of the overall pie.)
For the more complex example with widgets and sprockets, there's absolutely diminishing marginal utility; I dealt with this in the passage starting "you might wonder what happens if productivity gains mean we're making more widgets than we strictly need. We'd probably see layoffs". And in general we can assume that goods will suffer much more sharply diminishing marginal utility than services.
That, in turn, is what drives the shift in employment from more productive industries (which is another way of saying "industries which don't need as many people any more to fill societal demands") to less productive industries.
If it helps, consider the earlier shift from agriculture to manufacturing: Around 50-150 years ago, we saw an incredible increase in productivity, a very very small increase in per capita food production, a very large drop in food prices, a very large drop in the % of income the average person spent on food, and the utter gutting of agricultural employment. New technology gave us the power to make 40 times more food than we were previously making, or the same amount of food with 1/40th the people. And at the big picture level, that freed up labour ended up in manufacturing, which had not yet had the same productivity increase.
That's a good example of what Baumol's theory would predict. Does that make sense to you? Does your understanding of the theory suggest it would predict something else?
> Did we get voluntarily make education "less productive" by increasing the decreasing the student:teacher ratio? Did we get better educational quality for it?
Yes, we did lower productivity. And no, not really. But keep in mind, that when we say productivity here we're talking about labour productivity, so increasing employment will (basically) always lower productivity.
Yes, but not to Baumol's theory. Your description sounds like an excellent summary of the labor theory of value. How does this connect to musicians making more money? Once all of those former farmers rush into manufacturing because the price equilibrium is reached for food, why doesn't that depress factory wages and thereby reduce the musician's leverage to demand a higher salary for staying a musician? I keep rereading Vox's summary and one on The Economist, but I'm not seeing the evidence that musicians' wages were incommensurate over the century analyzed, or confirmation that other measures of music productivity like audience members reached per year remained fixed. If the musician's general relative economic class is unchanged, why invoke opportunity cost of other labor sectors as a factor? Or, if the general public has higher wages and the demand for music is inelastic, they can be stretched more for it, why not suppose the artists know their talent pool is limited and leverage that to increase their cut? I thank you very much for taking the time to try to explain this to me, in case I've inadvertently gotten you frustrated, and I blame myself for this apparent mental blind spot I can't shake.
One problem I've always had when stumbling upon the cost disease theory is my sense that real wages not only "should" rise due to productivity gains, but also due to the wages of alternative occupations.
If a person has the intellect / drive / grit / possibility etc. to pursue either an occupation as a university professor (where the productivity gain has been low) or a high end engineering job (with high productivity gains), it's obvious that if there's demand for university professors, universities will have to somewhat compete with the engineering firms for labour.
I guess the situation is somewhat different for the arts, but still, if the real wages had been constant since the 1600s I guess very few would pursue a career as professional classical pianists if it would be impossible to make more than minimum wage (even at the topmost levels).
That is exactly the point of the theory. Wages increase roughly at parity for equally skilled workers, regardless of their productivity, since otherwise people wouldn't choose to go into careers that under-pay. While costs decrease for goods and services that benefit from productivity increases, they stay the same for goods and services where productivity doesn't improve. Since all other costs are dropping, this has the effect of making prices go UP for hands-on services.
Most of this seems fairly uncontroversial. This, however
> But a lot of service workers are doing jobs that are unlikely to ever be fully automated. Nobody wants a robot for a teacher or a nanny, for example. And even if we get software with advanced diagnostic capabilities, patients are still going to want doctors to explain the recommendations and nurses to provide hands-on care.
Maybe the author is taking a very short term interpretation of "ever" but I don't know why these jobs are unlikely to be automated of general AI arrives.
We're social creatures. We like interacting with other people more than we like interacting with machines, even if machines are more efficient in a narrow technical sense. I think this is especially clear for nannies--the main job of the nanny is to cuddle and interact with the baby in ways that only a human being can do.
There are plenty of fitness videos and apps, but people still pay a lot of money for in-person fitness classes. People wouldn't stop going to Starbucks even if someone invented a vending machine that could produce coffee that tastes identical for half the cost.
> We like interacting with other people more than we like interacting with machines
I remember a '70s article in which Bank of America, I think, said that they wouldn't be rolling out ATMs like one of their competitors because customers prefer interacting with a warm friendly pretty teller, not a machine.
We know how the ATM vs human teller preference worked out. (I even have to stick the word "human" in there for clarity because "teller" is almost an archaic word.)
> People wouldn't stop going to Starbucks
People go to Starbucks or similar coffeehouses for many reasons. Better coffee than at home, chance to get away from home or the office, solitude (yes, ironically, you can have more privacy or alone time in a Starbucks than at home or office).
I figure that a Starbucks that was completely automated with no employees and little social interaction, but otherwise identical to a normal Starbucks, would still be quite popular.
Starbucks has these machines and I find it awesome to get my coffee and even chilled latte there for 2,50 instead of 5 or 6 $
The machines taking our jobs is already happening and it is due to our human nature that we are not seeing it: We extrapolate the past into the future and we are seeing physical robots
As some of you (technical people) can better imagine it is and will be much more 'just' software (but a journalist can't that well)
The "reduced" demand for nannys is 'kids playing with their iPhones' and it is already here and going stronger
What will we pay people for in the future? That they are likeable, sympathetic and because 'we like to "interact" with them'?
Seems like this would split the world into very hard skilled people programming the machines and a lot of people catering "what's left" (on the human side, that machines can't do / didn't replace yet) to those
Luckily I am a software developer
"Doctors to explain the recommendations": You could also say "Selling the pills of the Pharma Industry with low regards to side effects"
"Nurses to provide hands-on care": I see "Human Issues" like: She's telling you all her problems she has at home and you have to listen
Tell me the Internet isn't better in quality (than "real random people") already?
I think it is important here to draw a distinction between equivalence without type conversion (eg, ===) vs. equivalence with type conversion (==).
While it may be the case that a robonanny can change a diaper or administer a bottle with a precision greater than or equal to a human human nanny (==), it is not the case that the robonanny understands viscerally what its connection to its charge is (===). Humans are a product of biological evolution and as such, there are certain features of our cognition that are very, very difficult to overcome, not least of which is our collective hangup on our own internal categories (ie, types). This is precisely the reason many people experience possiveness, nostalgia, a fear of the "other", etc., and I believe it also goes a long way toward explaining why so many folks speak of interacting with robots as if they were human as 'icky' or otherwise suboptimal.
That said, I personally believe that we are doing ourselves a disservice by dismissing the humanistic value of providing, on a cultural/societal level, a means for most of us to keep ourselves occupied, fed, and sheltered, regardless of whatever impact such activity may or may not have on the finances of giant corporations.
You must live in a world where people don't make coffee at home due to cost considerations.
The main point of a nanny is to have someone else deal with your kid so that you don't have to.
Machines will almost certainly provide a worse experience than a human at the beginning, but the lower price expands the market significantly to those who could not afford those things before.
Teachers probably won't get replaced for the longest time since they're not paid out of your pocket.
I agree. I think robot nannies will become pretty commonplace as many of the people who can't afford human nannies take advantage of the cheaper mechanical option.
I was probably projecting, but I'd say day care, in a home or a centre, is a financial necessity, and a nanny is when you have too much money or too many children :)
But the larger point I was trying to make is that most people who need child care are looking for someone to make sure it doesn't die while they're away, not provide some special level of care.
A robot nanny might not be as good as the best humans, but it will be much better than the worst, and people don't like to expose their children to avoidable risk. It's like how I'll eat at McDonalds if I travel somewhere new instead of a possibly superior local restaurant. Consistent mediocrity can beat unreliable quality.
This seems worth dividing into a couple of concerns. One is: Can it provide an accurate enough simulation for the baby? I think we would definitely get there with advanced enough AI, since the level of sophistication needed for cuddling and interacting with babies is lower than, say, having an intelligent conversation.
The other concern is that even if the simulation is good enough, there's still an ick factor to "tricking" a baby into forming an emotional connection with a machine that doesn't have real human feelings.
If we build caretaker robots with software that closely models the internal function of human brains, to the point that we're satisfied that their outward behavior is accompanied by complex and human-like feelings of caring, that problem goes away too. Whether that's even possible is more of a philosophy of mind question; depends on whether you think internal experience depends on the exact nature of the physical substrate.
(Of course if it is possible, there's yet another layer there with the ethical issues of designing a slave race of beings with real feelings, see e.g. Robin Hanson's ideas about "ems.")
Maybe nanny is a bad example. I'd pay good money to get impartial therapy from an AI (guaranteed non-abusive) and then when it comes to solving the problems of humanity in general, a sufficiently smart AI seems like our best bet, on the basis that human nature is yet to be fully understood by humans after centuries of trying...
Because teachers and nannys are mechanisms of cultural reproduction. We could give those jobs to AI, but they're also some of the few jobs that are actually interesting in an economy where we have no material need for work.
http://www.epi.org/productivity-pay-gap/