NOTE: The Growth Economics Blog has moved sites. Click here to find this post at the new site.
Nouriel Roubini posted an article titled “Where Will All the Workers Go?”. A few pulls:
“The risk is that robotics and automation will displace workers in blue-collar manufacturing jobs before the dust of the Third Industrial Revolution settles.”
“But, unless the proper policies to nurture job growth are put in place, it remains uncertain whether demand for labor will continue to grow as technology marches forward.”
“Even that may not be sufficient, in which case it will become necessary to provide permanent income support to those whose jobs are displaced by software and machines.”
The worry here is that technology will replace certain jobs (particularly goods-producing jobs) and that there will literally be nothing for those people to do. They will presumably exit the labor market completely and possibly need permanent income support.
Let’s quickly deal with the “lump of labor” fallacy sitting behind this. Technology reduces the demand for labor in some industries, so fewer workers are employed there. Which raises the supply of labor in all the other industries. For that supply shock to generate no other employment you have to assume that the $15 trillion dollar a year U.S. economy is so rigidly inflexible that it has a definitely fixed set of jobs that can be filled. That’s ridiculous.
To a rough approximation, just about the exact same number of people work in goods-producing industries in 2013 (19 million) as did in 1950 (17 million). And yet somehow the rest of us have figured out what to do with ourselves in the interim. Between 1950 and 2013 the U.S. economy expanded from 28 million service jobs to 117 million service jobs (All stats from the BLS). You think that somehow it won’t be able to figure out what to do with more workers that are displaced by technology? We’ve been creating new kinds of jobs for two hundred years.
So let’s ignore the phantom worry that tens of millions workers suddenly find themselves completely at a loss to find work. The economy is going to find something for these people to do. The question is what kind of jobs these will be.
Will they be “bad jobs”? McJobs at retail outlets, wearing a nametag? These aren’t “good jobs”, real jobs. Making “stuff” is a real job, not some made-up bullshit service job.
We can worry about the quality of jobs, but the mistake here is to confound “good jobs” with manufacturing or goods-producing jobs. Manufacturing jobs are not inherently “good jobs”. There is nothing magic about repetitively assembling parts together. You think the people at Foxconn have good jobs? There is no greater dignity to manufacturing than to providing a service. Cops produce no goods. Nurses produce no goods. Teachers produce no goods.
Manufacturing jobs were historically “good jobs” because they came with benefits that were not found in other industries. Those benefits – job security, health care, regular raises – have nothing to do with the dignity of “real work” and lots to do with manufacturing being an industry that is conducive to unionization. The same scale economies that make gigantic factories productive also make them relatively easy places to organize. They have lots of workers collected in a single place, with definitive safety issues to address, and an ownership that can be hurt deeply by shutting down the cash flow they need to pay off debt. To beat home the point, consider that what we consider “good” service jobs – teacher, cop – are also heavily unionized. Public employees, no less.
If you want people to get “good jobs” – particularly those displaced by technology – then work to reverse the loss of labor’s negotiating power relative to ownership. Raise minimum wages. Alleviate the difficulty in unionizing service workers.
You want to smooth the transition for people who are displaced, and help them move into new industries? Great. Let’s have a discussion about our optimal level of social insurance and support for training and education. But the sectors people leave or eventually enter are irrelevant to that.
You want to worry about downward wage pressure as the demand for labor falls? Great. Worry about that. See the above point about raising labor’s negotiating power relative to ownership.
Hoping or trying to recreate the employment structure of 1950 is stupid. We don’t need that many people to assemble stuff together any more because we are so freaking good at it now. The expansion of service employment isn’t some kind of historical mistake we need to reverse.
Any job can be a “good job” if the worker and employer can coordinate on a good equilibrium. Costco coordinates on a high-wage, high-benefit, high-effort, low-turnover equilibrium. Sam’s Club coordinates on a low-wage, low-benefit, low-effort, high-turnover equilibrium. Both companies make money, but one provides better jobs than the other. So as technology continues to displace workers, think about how to get *all* companies to coordinate on the “good” equilibrium rather than pining for lost days of manly steelworkers or making the silly presumption that we will literally run out of things to do.
The Growth Economics Blog 
I lived and worked in northwest Indiana (teaching economics at Indiana University Northwest) from 1987 to 2012 (when I retired). One of the major issues facing “the region” (as it’s known in Indiana) was the collapse in employment of the steel industry. Employment fell from about 70,000 in the late 1970s to around 35,000 by 1984, and never recovered. It’s currently around 16,000 (and the steel industry is producing more steel in northwest Indiana than ever). A clear structural shock to the economy.
The reaction was:
(a) Steel recover, all we have to do is wait. (One of my functions was to tell people that the jobs were gone and were never coming back, which did not make me very popular).
(b) The jobs were still there, they’d just been out-sourced to stand-alone warehousing, or maintenance, or construction firms. (Another of my functions was to point out that employment gains in those sectors were very small.)
(c) OHMIGOD what do we do now?
The workers who lost their jobs were generally older–steel had not been hiring new workers for years. Retraining opportunities were not particularly abundant (a failure of state policy–and federal government policy for that matter). And for workers age 50 and older, retraining often seemed to be not a particularly attractive alternative.
Many of them left. Others were unemployed for a long time, and when they finally found jobs, were working in the growing sectors–retail, food service, health care (as maintenance or cleaning staff, often) at wages about half their prior level (or less). Total employment in the region did not regain its late 1970s level until the mid 990s.
I began studying economics in the 1960s, which was another era during which there was a great deal of discussion of the possibility/likelihood of long-term, structural disemployment. Which obviously did not happen. I still believe that long-term, structural disemployment is not the most likely outcome. But I now believe it will take much more forceful action to maintain full employment and to provide sufficient, and sufficiently attractive, retraining opportunities. And, frankly, the current makeup of Congress does not make me optimistic that appropriate policies will be forthcoming.
I hear you that reasonable ways to deal with disruption seem unlikely, even if more necessary than ever. My point would be that as we slowly struggle to get those policies, we shouldnt presume that only one kind of job counts when finding someone a good job. Lets make all the jobs good jobs.
I’m with you there. But the people who have lost their jobs also have to buy into that.
The concern is that technological change will lower the marginal product of unskilled labor until it is below the minimum wage or the subsistence level. For this it is necessary that the elasticity of substitution between capital and labor is greater than unity. Up till now we have lived in a world where the elasticity of substitution is between capital and labor is less than unity and using more capital with workers raises their marginal product. The concern is that this has changed now.
The more I have thiught about this, the more convinced I am that the EOS is less than one, and way less than one for lots of services. The reason is that for many services, I want time. I want a one hour massage, and it is impossible to substitute anything for that hour of time. In that sense, the economy can keep finding jobs for people, by putting them to work in jobs that are defined by their provision of time.
I’ve thought about that, but say 20% of the population is skilled and/or owns a lot of robots/capital. So they have the money to pay for a masseur, a personal trainer, a housekeeper. Even if they hire the equivalent of one full-time person, that still only gives jobs to 20 of the other 80 percent.
And that’s optimistic to start with, because studies are showing that people become comfortable with pleasant robots/computers providing services very quickly. A robot masseur can provide as good or better a massage than a person, but at $2 per massage instead of the $20-30 needed to give that person subsistence; how many of the 20% will pay the huge extra for the person, same for the housekeeper, and even the personal trainer (and how many hours is that). And robots can be way more convenient. An ATM is 24/7/365, so is a robot massager.
In the medium run, I don’t see much hope without obtaining a pretty high level of education and skill. For more see here:
http://carolabinder.blogspot.com/2014/03/guess-post-second-machine-age-book.html
I think the mistake is to think that personal services are something only 20% of the population would/will/can buy. If you do see displacement of people then those services will be widely available, and goods will be incredibly cheap. So most people will be able be buying services from each other. I don’t think that *necessarily* the shift to personal services means that we all work for the wealthy.
The other 80% can’t buy services from each other if they have no job, no earning power. If the makers of goods, the owners of the capital/robots and the high skilled, can’t find a use for them, then they won’t have the goods needed to exchange for services from other members of the 80%.
Nobel prize winning economist Jan Tinbergen said that inequality was a “race between education and technology”. It’s just that in the past education kept up; there was a huge public investment. But today, there’s so much billionaire money going into promoting the idea that any government is bad (except when it’s for the rich), and the amount of education and skill needed is really large and hard to get. We’re not keeping, up and soon technology may really take off, with little increase in the education and skill level to match, leaving a large percentage without the ability to command minimum wage in the marketplace.
Horses didn’t just find new uses when technology advanced. The wage, of course, adjusted so that they still could find work, but to below the cost of their subsistence for the vast majority of them.
Except that people buying stuff from one another is exactly what an economy *is*. So yes, a bunch of people can buy services from one another, and that can be the basis on plenty of economic exchange. On the logic you used, there are a fixed number of tasks to possibly accomplish. I dont buy that.
The 80% can buy and sell services amongst each other, as they all have that to offer, but then how do they get goods, that none of them have to offer, and don’t have the earning power to get? Two hungry people can exchange massages and house cleaning, but neither will get food, unless they can get hired by those who have and produce the food.
The same way they get food now? By exchanging services for goods? The whole displacement thing occurs because tech makes it ridiculously cheap to make goods, so the relative price if goods is likely to fall.
The amount of services that the wealthy will need — that robots can’t provide for less than minimum wage per hour — may be way less than the supply of unskilled. There’s only so many hours of personal training needed by the wealthy, and it’s not many, even if the wage drops to minimum (if it isn’t already, including everything); there’s only so much house cleaning that they will need (that robots can’t do), even if the wage drops to minimum (and it’s pretty close already),… You may run out of things pretty quick that robots can’t do for less than minimum wage (or less than a dollar per hour).
The vast majority of horses didn’t find all these services they could do in exchange for enough food for subsistence. Even though some did, recreational rides, racing,…, it wasn’t enough for the vast majority of horses that existed before the tech revolution.
My response is again that personal services are not just something that wealthy people use. Lots of unskilled personal service types provide unskilled personal service to other unskilled personal servicers. The fact that it is an unskilled personal service does not, by itself, mean that it will be un-valuable.
“I want a one hour massage, and it is impossible to substitute anything for that hour of time.”
“The concern is that technological change will lower the marginal product of unskilled labor until it is below the minimum wage or the subsistence level.
Neither of you are discovering anything new (http://en.wikipedia.org/wiki/Baumol%27s_cost_disease).
First of all, technology *is* changing the way we deliver *services* like medicine, education, housing, and art. We have diagnostic medical programs, and computer CGI. Labor productivity is going up in those areas, just not as fast. If it went up faster, health care and education would be cheaper.
When (not if) someone invents a computer-human brain interface with the ability to download skills, the whole education system is going to be turned on its heels. OMG! professors! out of work!
The jobs people do in 2319 are the ones for which labor productivity has improved the least. Personal services and entertainment, maybe.
But, there will also be an *enormous* economic surplus. People will not have to work to feed themselves, clothe themselves, put a roof over their head, learn algebra (just download it!), or to diagnose illness. Or to maintain and build more robots.
People value their leisure time! If they don’t *have* to work, why do they need to? Will we decry the “dependency culture” that has arisen from robots, because most of what people all the time do is fly around the world and see new things? You mean I can spend a year in Rome, a year in Hong Kong, year travelling the Chinese countryside… wow, you will have to pay me a lot to get off my duff and pull me from my vacation since their is no real need for me to work…
… wow, everyone has time to participate in a real historical re-enactment of WWII. sans, death of course. see here: http://www.thisiscolossal.com/2013/09/the-fallen-9000/
When 20 billion people have so much free time, the possibilities for “art” are endless!
Well, except one thing. If everyone lives a life of leisure, I foresee a need for congestion pricing, so prevent all 20 billion of us from choking into Rome on Easter to see a reenactment of the 1st Easter, we may have to charge people.
damn, I may need to rearrange my schedule because I really wanted to do it this year! Or I could give 10 massages and pay the fee. hmmm.How badly do I want to go…
Baumol is a great reference here. The only difference in the current debate is that people are worried about workers literally having *nothing* to do.
But this, as you say, is not a bad thing. A perfectly fair response to tech. change is to reduce inputs, not raise output. The big question is how we manage to get everyone to coordinate on lower work hours. (Always an incentive to cheat to get a little ahead by working 12 hours, not 11 per week).
I hope you’re right about Rome and Easter. Means I can get a table at a restaurant.
Dwb,
I really enjoyed this comment.
Dvollrath,
Could you expand on why you need to coordinate around shorter hours? Extra hours are just an inefficient way to increase productivity. It isn’t apparent to me why I care if productivity is doubled via technology or someone else working an extra shift. I can see why THEY may care, but if they care, I don’t need to coordinate with them.
Are you suggesting employers will prefer the person with double hours? If so, I can see how coordination will be necessary. Not sure I trust our ability to nudge in this direction though either.
Right, employers may prefer people willing to work more hours (maybe fixed costs per worker). So how do we agree not to “cheat”? I agree, not obvious how you make this happen. Just suggesting that without doing it, we might not get the 3-hour day any time soon.
Of course, the question is what these new jobs will pay, and for most of the low skilled that might be less than minimum wage, if the were to get a job. The serious question is will advanced revolutionary computers/robots be able to do so much that a low skilled person can do, at the equivalent of less than minimum wage, or far less, that sure you could find something left for them to do, but nothing that people would pay anything close to minimum wage for.
I think you are framing a straw man assumption when you write:
“For that supply shock to generate no other employment you have to assume that the $15 trillion dollar a year U.S. economy is so rigidly inflexible that it has a definitely fixed set of jobs that can be filled. That’s ridiculous.”
I would re-write as:
“For that supply shock to generate no significant amount of other employment, you have to assume that the $15 trillion dollar a year U.S. economy is so flexible that automated systems—both pure IT and robotics—will be widely adopted by firms as soon as their amortized acquisition and maintenance costs drop below any wage that would prevent a full-time employee from starving, and that initial successes in this trend will provide the competitive advantage required to finance its development to its logical conclusion.”
I think you are being wildly optimistic to believe that there is any area that won’t be so affected, and between Baxter (http://www.rethinkrobotics.com/baxter/), self-driving cars, and domain-specific expert systems (think medical diagonstics), I see the trend lines on my side. Intelligent people can disagree, but those who paint the continuation of existing trends as impossible tend to be on the losing side of history.
If you think some inherent limit exists, please be explicit.
Else: prepare for permanent, wide-spread income support.
It’s not so much a limit on automation – I’m sure we can program a robot/computer to do pretty much anything. The limit is that we have demand for things that are literally impossible to automate out of existence. If I want Bryan Caswell (Houston-area chef) to cook me meal, then I want Bryan Caswell to cook me a meal. You cannot automate that. Could his kitchen be more automated, and perhaps need fewer workers? Sure. But there are services that we demand that are purely based on the attention of another particular human. By definition, those cannot be replaced. And I think that the economy can expand to offer more of those kinds of services.
Easily? Probably not. Without costs of disruption? Certainly not. But just like we managed to handle the swarm of millions of farm workers leaving for cities in the early 20th century, we’ll handle this eventually.
> work to reverse the loss of labor’s negotiating power relative to ownership. <
THIS is the key to restoring the middle class.
Unfortunately, I am 100% confident that nobody reading this will live to see meaningful action on this (at least in the US).
I think part of the issue is that it isn’t immediately obvious what the right policy/choice/action/argument is to make to improve labor’s negotiating position. Leave aside the political issue – what one concrete policy change should we make? It’s not easy to answer, and that it part of why it’s hard to make the case.
sigh. There is no such thing as a “good job.” Seriously, if you were going to do it all on your own without getting paid, it would be called “vacation” or “retirement.”
A job is merely something someone is willing to pay you to do. If it is not worthwhile, no job.
When robots build themselves, make food, build houses, and take care of us, probably people will have a lot of free time and will pay each other for entertainment. Or maybe they will write inane stuff on a blog so people kill time making comments.
There are jobs that have less uncertainty with respect to future employment, more autonomy at work, and higher benefits. Even if it sucks to go do them, they are “good jobs” from one perspective. There’s a reason people want those jobs over other ones.
Of course there is differential supply and demand for different jobs! Thank you captain obvious!
As for job uncertainty, You are confusing supply and demand.
In the case of supply, the lower your reservation wage (the lower the wage at which you are willing to work), the *more* job certainty you have, because it’s more likely someone will accept your wage offer. In the case of demand, it’s the opposite, the more you are willing to pay a worker, the more likely the job will get filled.
If the demand for labor is inelastic (which it is short term) job certainty is *inversely* correlated to wage (including benefits like autonomy). The higher my reservation wage, the less likely to find a job.
“McJobs” are not bullshit jobs, Capt Obvious. Busy people want to have stuff done for them: Eat, clean their house, scrub their toilets, do their laundry, mow their lawn, and prepare and preserve their dead relatives so they can take selfies with the deceased. They want to go through a drive-through and have someone flip a burger or serve them donuts and coffee while their child is screaming in the back. This has been going on for eons. I guarantee you, in the future of robots, some of these jobs people will want *other people* to do.
It is not the *demand* for these jobs that is uncertain. Burgers need to be flipped. People have questions about their sweater color… like do I look fat in this. They want a “Nametag” to reassure them and like dealing with people. These jobs are “uncertain” because there are too many people willing to do it at market wage.
If you are willing to mow lawns in Maryland for $1/hr (and the law allowed it), I guarantee you will have a steady supply of work. You will probably raise your price even to choke it all off. If you charge $100/hr, you may have trouble finding work. In North Dakota, where there are not enough people, “McJobs” pay much more (>$15/hr) and job certainty is 100%. See how that works?
Of course people want to supply their labor at higher wages! duh! “Higher skilled jobs” are paid more because there are barriers to entry. Of course everyone likes to argue and be a lawyer, we don’t allow that in the U.S. Or, accountant, petroleum engineer, doctor, or most other hot jobs.
Perhaps you should be asking, why are there so many barriers to entry in the job market, who do they really protect, and what we can be doing to improve the appalling high school graduation rate in NYC public schools (there are 0000s of kids who drop out with no diploma, and “diploma” is one of those barriers to entry).
On the other hand, “dope dealing” has no barriers to entry, and it turns out pays pretty handsomely once you climb the ladder.
Stop worrying about good jobs, bad jobs, McJobs, and bullshit jobs, and worry about breaking down barriers to entry into the labor market.
Thanks for reading, I guess. Let’s skip past the blather, and get to your good point. Eliminating barriers to entry (licensing costs, artificial exams/tests requirements, etc..) would be a great idea. There is a lot of evidence that these barriers have reduced labor mobility between jobs, so I’ll happily sign on to a plan to remove those barriers.
How do you think the trend towards manufacturing automation will impact developing economies? Rodrik keeps making the point that the structural transformation through manufacturing doesn’t seem to be happening anymore. How viable are high productivity service sectors in low-income countries? It seems like the educational gap is much tougher to cover for the service sector than it was for the manufacturing sector in the past.
Good question. I haven’t totally gotten my head around Rodrik’s argument, but there is something there. The optimistic take is that they can leap-frog right to a high-service economy. The pessimistic take is that you have to somehow “learn” your way through manufacturing to get to high income. I feel like that pessimistic take is too rigid, but I can’t totally sell myself on the optimistic one either.
“That’s ridiculous.” Really? So you don’t believe there has been and can be such a thing as severe and/or prolonged unemployment in America, and you don’t believe that other nations have gone for decades or even centuries with youth unemployment in the 30-50% range? And are jobless recoveries, 30-year trade deficits, sticky wages, and shrinking labor share of income other things that don’t occur in nature?
Really. Those other things you mention are all widely attributed to policy problems/choices, but they are not generally problems that arise because of technological change. 30% youth unemployment in Europe is not because technology took all the jobs, its because the sclerotic labor market removes all incentives to hire new workers.
Even if robots do take over, that won’t necessarily be bad. I for one have a huge preference for leisure and would enjoy if the 3-hour working day Keynes was talking about became reality.
Of course, we’d probably have to rethink how to redistribute resources in a manner that allows people to live above subsistence levels, but that’s a whole different issue.
I think that’s the big question. Robots come along, and conceptually we all could work 3 hour days. How do we get everyone to mutually agree to that? Because individual workers would have an incentive to deviate to keep jobs (I’ll work 4 hours, not like that lazy jerk Bob!). But we saw things like this in the past with introduction of 40-hour work week. So it probably involves some kind of policy change to enforce the mutually agreed lower hours.
I’m with you, though, bring on the short days.
Pingback: Automation and workers | occasional links & commentary
I agree with not confounding good jobs with goods producing jobs, but then you go on to confound good jobs with union jobs.
Implicit in your argument is that the wages and benefits and working conditions in manufacturing are a result of bargaining as opposed to marginal productivity and competition. Although I won’t argue that collective bargaining is irrelevant, productivity and competition are the cars driving this phenomenon and unions are more akin to the floor mats and detailing packages on the car.
You then give a partial list of “good” unionized service jobs. Notable is how you left out claims adjusters, supervisors, programmers, human resource interviewers, market analysts, accountants, etc. You simply rattled off a list of jobs which met your definition of good then presented that as evidence that the two were equivalent. They aren’t.
Indeed, teachers and cops are probably not considered pleasant jobs with especially amenable working conditions for those in these fields. Furthermore, they are in two of the most widely condemned industries in terms of productivity. In both fields we see rising consumer/public dissatisfaction with the results compared to the costs. No need to link to the trends of education doubling in cost over the past generation with almost no gains in outcomes. For cops we need look no further than Ferguson and NYC. Both look more like fields captured by rent seekers at the expense of the public at large. For you to present these as the model for good jobs is certainly questionable. I am surprise you didn’t include health care unions, or the California “public services” cartels.
Then, oddly, you go on to use an example of a largely non-union company, Costco, to drive your point home for aligning on a good equilibrium. Putting aside the fact that this kind of contradicts your collective bargaining argument, implicit in this line of thought is that a high wage/benefit strategy is superior for the employee (and ultimately the consumer) than a low wage/ benefit strategy. Sorry, this is only true for the employees if you hold all potential employees as equivalent in all ways. In the real world, less experienced and capable employees compete with more experienced and capable ones with different values and needs. Making all jobs high paying with great benefits just handicaps those with less abilities from ever getting a job.
Said another way, low skilled, less experienced employees and people with cognitive and behavioral issues need opportunities too.
I could go on here, but in brief you are confusing cause and effect and ignoring secondary and unseen effects. Standards of living adjusted for cost of living are lower in those states more easy to unionize coercively, not higher. Minimum wage and mandatory benefit increases make labor markets less efficient and harm the less skilled and capable relative to their true competition — which is other potential employees (employees cooperate with Wal-Mart negotiating only on the terms of the reciprocity, they compete with other employees and potential employees). And most importantly your definition of great jobs are ones which ignore any semblance of productivity or consumer benefit.
In summary,
1) good jobs are not the same as jobs with collective bargaining
2) the big driver here is productivity not bargaining
3) you are ignoring secondary effects and looking at it from the view of the employee rather than that of the consumer/employee holistically
4) you are confusing cooperation and competition
5) you are reversing cause and effect and trying to raise living standards by fiat
I’m with you – and the post comes off as more pro-union than I’d like in retrospect. The point is really that one of the reasons we think of manuf. jobs as “good jobs” is because they did have union power to extract concessions. This definitely has costs: outsiders are excluded, firms may not expand, etc.. etc..
But the general point remains, I think. Which is that if you want to get people “good jobs”, then you need to do something to get firms/workers to coordinate on the “good job” equilibrium. Unions were one way to do that, perhaps very bluntly. Big question – why/how did Costco get where it did? Is that something that is replicable? Not saying it has to be a policy change – is it just something that workers need to learn about in order to negotiate with?
Well, I recall Megan McArdle wrote a column a few months ago on the Costco issue.
http://www.bloombergview.com/articles/2013-08-27/why-walmart-will-never-pay-like-costco
The good job equilibrium is one of high marginal productivity. The negotiation power comes from their potential productivity along with demand and competition for this productivity.
I fear most of your original prescriptions would actually backfire and lead not to the good equilibrium as much as the elimination by fiat of what are being viewed by you as a “bad equilibrium” and by the Walmart employee as a “job.”
The heart of the matter seems to me to be the dynamic responsiveness of the system. Certainly someone can come up with jobs, but the question is how long does this transition take. As technology accelerates at a faster rate, will it eliminate jobs faster than the rate at which they are created and the labor market can refill? And I don’t think education is the key (though super important). I think the key is the pace of technology and the corresponding flexibility and efficiency of labor markets. Tech is changing faster and labor markets are getting less efficient. Bad combo.
Pingback: Technology and “Good Jobs” | The Growth Economics Blog
Pingback: Techno-neutrality | The Growth Economics Blog
Pingback: Interesting (?) thoughts on wage growth and good jobs | Avoiding Citation