Friday, February 28, 2014

The jobless economy

I was recently interviewed by Tom Ashbrook for NPR's "On Point" about the "jobless economy" - basically, the "rise of the robots" that everyone is talking about. Erik Brynjolfsson and Andy McAfee, authors of the new book The Second Machine Age, were also on the program. You can listen to the audio here.

Erik and Andy had very coherent and well-prepared points, which is not surprising given that they've just written a book on the subject. I haven't yet formed a single dominant thesis about the "rise of the robots", so my thoughts were separated into a number of distinct points. Those points were:

1. Technology has not been as important a force behind inequality and unemployment during the last 15 years as many people think; up til now, the much bigger story has been globalization (especially China). This is supported by Mike Elsby's research on the subject. But going forward, the "rise of the robots" may be a much bigger worry, and we need to think about how to respond to it.

2. It's not yet clear if human labor will find a way to remain valuable en masse in the age of automation and the digital economy. That ended up happening in the Industrial Revolution, but there is no guarantee that it will happen again. We need to be prepared for the possibility that large numbers of humans are rendered nearly obsolete by new technology.

3. In the short term (~5 yrs), infrastructure spending can boost our economy and help people get more jobs and higher wages. But that opportunity will be played out and will not repeat itself.

4. In the medium term (~20 yrs), our culture of work and the dignity that paid employment provides can be partially preserved through a program of wage subsidies, which are similar to the EITC but easier to implement and will probably have a more positive psychological impact.

5. In the longer term (>20 yrs.), if technology continues to make more and more and more humans obsolete, we need to do two things. First, we need to find a way to distribute income more widely. Redistributive taxation and "basic income" is useful but it has its limits. Finding a way to redistribute capital income - by helping Americans to be small business owners en masse, and/or by finding a way to democratize ownership of most companies - will be essential. The second thing we will need to do is to transition from our traditional culture of work to a culture that values humans for their interpersonal relationships and self-expressive creative output - two things that will never be obsolete.

But remember, all this is contingent on the progress of technology, and that is something that economists and futurists alike have always found devilishly hard to predict.


  1. Anonymous12:34 PM

    If we convert capital into spending ala the inheritance tax, we'll end up with a lot less capital. If we use it to create a sovereign wealth fund, we'll politicize investment, which should be really awesome. If we redistribute ownership (e.g., as a "birthday" savings account for newborns) we'll do better.

    "Transitioning a culture" is easy to say, but has never(?) been done in a conscious fashion. Personally I'd prefer installing the value of "competence" in whatever domain, in that it is a lot closer to "work" than "self-expressive creative output". That would include, say, Wikipedia editors and search-and-rescue volunteers, without excluding that artists and "friends"(?) that you want to reify.

  2. JeffD1:09 PM

    To your #2, I gather Piketty's new book has something to say about the degree to which the Industrial Revolution allowed labor to remain valuable. Absent the huge shocks of WWI-Great Depression-WWII it seems like labor may not have attained the income share it did in the 50s-70s. I'm only through the first chapter thus far though, so obviously I may be mistaken.

  3. Great article from Martin Wolf on this (paywalled unfortunately):,Authorised=false.html?

  4. Anonymous1:16 PM

    Re: #1, I've been hearing for two decades that the decline in secure, well-paying, middle-class manufacturing employment is not due to China, but due to technology / the increasing automation of manufacturing. This seems inconsistent with your claim. Either one of these is incorrect or something else that is very unintuitive is going on (technology's effect on manufacturing employment is washed out by China's effect on employment in the fast food sector?). It might be nice to explain this point more fully.

    1. This comment has been removed by the author.

    2. You say: "the decline in secure, well-paying, middle-class manufacturing employment is not due to China, but due to technology"

      Noah says: "the much bigger story has been globalization (especially China). This is supported by Mike Elsby's research on the subject. But going forward, the "rise of the robots" may be a much bigger worry"

      So yes, there's an inconsistency - your beliefs versus current research. According to current research, the problem has been cheap labor abroad. The future problem is robots replacing that cheap labor.

      I think what you've been hearing from the last two decades is wrong.

  5. Only about halfway through the show, so apologies if this comes up later and is better addressed.

    The issue that seems to bother Tom the most is the availability of a 'job' for people, which sounds like traditional, salaried labor. One of the lessons of WhatsApp, among other such stories, is that high-value enterprises require far less salaried labor to meet large-scale demand of their products or services. So, perhaps the issue here is first trying to apply that culture of work onto the future, where instead of salaried labor we might have something more akin to a freelance or a direct-to-consumer type of labor structure dominated by an entrepreneurial, or be-your-own-boss type of structure. One element of this that comes to mind is turning hobbies or 'work that you love' into a side business, while also doing some traditional salaried work (cultural transitional period). The growing strength of the individual to compete with traditional or corporate businesses is already seen in tech, media, entertainment, etc, which were all discussed on the program as the most transformed industries.

    Instead of discussing how to defend large-scale, traditional, salaried labor models, perhaps a more fruitful discussion is one focused on policy, namely, how to make it easier for individuals to compete going forward, either by limiting corporations from manipulating policy to make such competition harder, or by reducing already high barriers to entry for individuals or small companies. One example that comes to mind are the regional licensing restrictions for occupations like interior designer, hair dresser, and the like that really grind Matt Yglesias's gears.

    One more off-hand note is that I would imagine a premium will be placed on 'man-made' or 'craft' work in the future, much as it is today. This is work is often regionalized (more immune to globalizing pressures) and at least partly dependent on human involvement. This work would fit into a 'skilled, non-automated' type of labor that appears to be where the future wage stability might be, while also offering a wage premium above the menial labor that is also immune to automation.

    1. Anonymous2:03 PM

      The rise of 3d printing and other automated, yet customizable manufacturing processes may make "craft" mean something very different. I can design a customized coffee cup/dining room table/artificial hip/car seat/etc. and manufacture it in volume with no labor other than refilling the ink wells...

    2. If the question is simply one of value-added (beyond comparable mass automated product), then perhaps what I meant to say is something relatively unique, personalized, or otherwise meaningful. A shirt that reads "Fight the Power," that is made in China by a multinational known for using child labor, for instance, will be valued quite differently from the same shirt produced by a local t-shirt company known for using high-quality local or fair-trade materials. I guess its not necessarily (or only) about non-automated work, but the spirit of the work itself, which I would consider the 'craft' or 'art' of work. This is probably too abstract a concept to be valid at the level of theorizing a post-job economy. However, we already value such differences as a culture. The loss of a job-dominated economy, where categories of work are silo'd into functions might shift to a growing niche labor and consumer market. Such a shift demands certain policy shifts as well as a more adaptive and creative workforce.

  6. Anonymous1:28 PM

    To me, Noah is too liberal somehow.
    He is the man who never provides viable and practical solutions. It has absolutely no problems, though. haha
    After all, he is an academic living in an ivory tower. haha
    Well, I think Americans tend to think too highly of the impact of technology while thinking too lowly of the powers of government.
    The process of replacing humans with robots would be more like a drawn-out war, I guess.
    To predict a future, you should look into the inner workings of human minds, something small business owners like me excel at.

    1. Anonymous4:59 AM

      Dunning-Kruger explains a lot more than just bitcoin.

    2. Anonymous10:17 PM

      I disagree, a drawn out war requires opposing forces. The only force that will oppose the robotic workforce will be the workers. Business owners will switch to robotic workers quickly due to the immediate economic benefits. If you owned a small food market and I offered you a way to increase your profits while lowering your costs and reducing your liability, would you not take this offer? Robots will cost around 1-4x the annual salary of an average employee yet could arguably last for decades, they can work 24 hours each and every day, they can be upgraded and improved over time, and they can produce a consistent and measurable quantity and quality of work. Let's say the business owner chooses to stand their ground based on their principles of providing jobs to human beings. Will their customers continue to shop there when the competition, that does use robotic workers, is charging an average of 20% less for arguably higher quality service / products?

      Money is going to be the driving force behind the adoption of robots, not a love for technology. I'm guessing we will see a wave of technically sophisticated companies dominate most old and established non-technical industries. It will be too attractive of an opportunity for robotics companies. I bet we see a group of robotics companies that cover a broad range of unrelated industries.

  7. The solution to high unemployment in the future will be a revitalized labor market for artisans, and space explorers.

  8. I'm still skeptical that we won't figure out new jobs for people to do, even if low pay becomes an issue (which it might). You never see the types of jobs that come with major technological and economic shifts, just like how somebody in the late 19th century would never have seen countless computer-related jobs popping up.

    Maybe we'll all be mechanics or robot shepherds, at least those of us who aren't providing services where having a human is the point. There was an interesting discussion thread over at Wired over unmanned ships, and an engine room engineer made the point that crews have basically shrunken to the point where automation to replace them would be enormously expensive and complex (i.e. prone to break down).

    1. Anonymous3:45 PM

      Since the 200 tech crash. the only significant job increases have been in health care and ed (mostly admin). Neither can continue, and in fact, productivity revolutions in both look imminent.

      Consider Amazon. Not only is it eating a bigger share of commerce every day, how long will it be before no human touches a good between the time it comes off a truck at a distribution warehouse and the time it goes back on a truck for delivery? How much longer after that before that same good goes untouched between the time its raw materials enter the factory and the moment when the drone drops it off at your front door?

      Do we turn warehouse workers and truck drivers into nurses?

    2. Maybe not nurses as we define them now, but why not more health care and administration workers? The larger and more complex an economy gets, the more management and administrative tasks are needed just to manage that complexity. And there's ample demand for health services, particularly as other forms of spending as a percentage of household budgets due to productivity gains.

  9. If only prices would fall so most everyone could afford the benefits. In the industrial revolution most of the value added was gained by consumers. Now with ever heightening IP protection and captive corporate boards, less is available to everyone else. In part the production of free products helps combat this but has still done little to reduce overall costs.

  10. The theory is that increasingly cheap robot production will drive down wages, but it seems like human labor should remain competitive at some price. If it's an economy-wide trend, then the prices of my consumption bundle will probably more or less go down under the same pressure as my wages, and in line with them. Yes? Seems like I should remain employable, willing to work, and no more impoverished than I am now, even if the price of my labor is under pure substitution pressure from automation. So how does this become a story of obsoletion and unemployment rather than straightforward income inequality?

    Your remedies do seem reasonable either way.

  11. Anonymous9:39 AM

    It feels to me like some kind of fallacy of composition is being committed here. If people are rendered obsolete and cannot earn wages to pay for and consume the output of the robots, then where is this output going? How can you have all these robots producing without people consuming? Who will be the buyers for the goods and services the robots (or their owners) are selling?

    It also feels to me that a massive market failure is being described. No matter how inefficient human labour is compared to robots, an efficient market should put both to work. It is not a choice between people working or robots working -- why can't we have both working? It is not like there is a fixed quantum of work that the robots can fill, leaving none for the people -- there is no "lump of labour". If the long-term outcome has lots of people unemployed -- i.e. resources left persistently idle -- and widespread unhappiness, then there is a something seriously wrong with the market rules.

    I think it more likely that wages and prices will fall until robot and human labour regain parity. The result will not be mass unemployment but deflation and inequality. The correct models to look at are slave-owning economies of the past and economies with very large proportions of immigrant guest workers today. These are economies with lots of "robots" which can do anything the local human workforce can do much more cheaply. I think that unemployment of the local workforce was not really a feature of these economies, but inequality and deflation were.

  12. Anonymous11:24 AM

    I take issue with this thesis that robots will make much of human work obsolete, for two reasons: 1. It has been disproven by recent history (last 200 years). 2. This view takes a very US-centric approach to things.

    To explore the second point a little bit, once we start to see mass unemployment in countries with better macro-economic management than the US (Australia, Canada, Nordic Europe), we should start treating the "rise of robots" theory of mass unemployment and wage stagnation more seriously. Otherwise, I think such a thesis placates us in terms of reforming our political economy...

  13. Anonymous12:50 PM

    I hate to disagree with someone smarter than me, but it's hard to imagine that in the long run machines will make people obsolete en masse.

    We have so many problems to solve and work on:

    - Cancer
    - Alzheimer's
    - AIDS
    - Renewable energy
    - Space travel
    - Sustainably ending food shortages
    - Economic growth in Africa
    - International financial stability
    - Advanced prosthetics
    - Etc etc etc etc etc etc etc etc etc etc

    All of these require tons of human input in all kinds of areas: law, finance, engineering, public policy, life sciences, and yes- artificial intelligence.

    As someone who has worked on a fair number of "real world" projects I believe that they require a lot of things that computers are unlikely to excel at in the near future:

    - Excellent communication skills
    - Learning new things from different domains quickly and figuring out how to apply the new knowledge
    - Using old concepts in novel ways
    - The ability to read between the lines of what people say
    - The ability to differentiate between explicit and implicit rules

    This fourth one is critical, IMO.

    Will it be easy to produce robots that know when to obey explicit rules and when to disobey them because they have implicit permissions to do so?

    This kind of stuff is crucial to negotiations, transactions, policy formulations, politics, IT projects, budgeting, etc. from what I've seen.

    Last thing- Watson can answer questions, but can he write research papers? If I ask him what the capital of Mozambique is, he would tell me in a second.

    But if I ask him whether we should litigate a 500-million-dollar pharmaceutical lawsuit in Dutch court, US court, or Swiss court, will he be able to tell me why and explain the pros and cons given the circumstances of my company, in a language tailored to my CEO's shoot-first-ask-questions-later personality?

    IMO the difficulty curve of making Watson go from answering Jeopardy questions to making him make nuances legal arguments is steeper than the newspapers would have you believe.

    1. Anonymous2:17 PM

      "IMO the difficulty curve of making Watson go from answering Jeopardy questions to making him make nuances legal arguments is steeper than the newspapers would have you believe."

      This would be much more relevant if most humans were currently employed making nuanced legal arguments. A cursory examination of reality will reveal the problem here.

    2. Anonymous4:15 PM

      Of course most people aren't lawyers.

      Do you really think that's what I am saying?

      There is a broader point here.

      Most people in the service industry (the majority in the developed world) are in fact engaged in doing the type of work that requires mental dexterity, good communication, flexibility, etc. which legal work is simply an example of.

      Whether you are planning a corporate retreat, taking care of the office plants, or arguing Supreme Court cases, you need to be able to respond to dynamic problems, explain things to others, understand ambiguous requirements, etc. at least some of the time.

      For example, the lady that waters our office plants recently found that some of them were dying. She asked around and found that the guys in those offices tended to keep their blinds closed.

      She kindly asked them to keep the blinds open and explained why. Some refused, so she moved those plants out of their offices.

      I doubt any robot would handle this very simple series of events well any time soon. It's also not likely to be economically profitable to create robots for every single real world challenge people can solve using the learning capabilities humans have evolved over several millions year of biological development.

      I would think it's obvious to anyone who has spent a considerable amount of time in service work that most jobs- even administrative and secretarial jobs- involve much more than algorithmic work that never changes.

      So the threat is to manufacturing and agricultural jobs, but that's not exactly a portrait of "mass unemployment".

      PS: Do you have much experience creating robots? It often seems that people who don't do a lot of work around automation are the ones who exaggerate how easy and fast it is to automate human tasks.

  14. Nice Info! Thanks to the tremendous popularity of the Norton software, it has found a way into the computers and laptops of millions of users all over the globe.

    Norton Tech Support | Mcafee Tech Support

  15. Anonymous7:14 PM

    I think we're assuming robots will replace human work entirely, which is nearly impossibile due to the sheer complexity of the task. How many kinds of robots must we invent to replace EVERY job available? It may be an entire job field itself !
    I can see some resemblances with the computer related fears in the old age. Computers haven't replaced our minds and souls, just the work we used to do without computers. We haven't become OBSOLETE because of machine computing, just like cars didn't make our feet useless. But we live a MUCH better lifes thanks to them.
    So, just imagine some robot doing something we classify as "top job", like medicine. We'll have a medicine robot, that can do exams, diagnoses, basic healings, even prescribe drugs. It's quite difficult it will replace a human doctor entirely, because it'll be too complex and costly. But it will save, say, half the time the doctor takes to do basic exams, diagnoses, healings, etc. So we'll have a doctor working half the hours, saving them for the serious cases, and leaving the routine work to the robot. On the customer side, you'll have an ultra-cheap, 100% efficient and reliable doctor available 24/7, for at least some of your health needs. A better world for both.
    Surely it'll not come without downsides (i.e. pollution), but I think mankind willl make itself essential at least fo another while.