The Limits Of The Digital Revolution: Why Our Washing Machines Won’t Go To The Moon

David Autor,  Professor of Economics at the Massachusetts Institute of Technology (MIT), joins Social Europe Editor-in-Chief Henning Meyer to discuss the impact of technological changes on the world of work and the wider economy. The discussion highlights why washing machines will not go to the moon any time soon and why the developing world might have more to fear from the digital revolution than rich countries.

David, thank you very much for joining us today to talk about the future of labour markets, the future of work, and what digitisation and technological progress in general mean for it. In your publications you mentioned that there are, as you see it, limits to the polarisation of labour markets that we currently see. What do you mean by that; where are these limits?

Sure. To clarify, I don’t mean that there are limits to technological progress or what will be accomplished; I just mean that over the last 20 years in particular, maybe even 30 years, one of the very important manifestations of computerisation has been the automation of routine tasks. So, taking things that people did that were cognitively intensive but followed well-understood procedures, like adding and subtracting, and filing and sorting, and processing information, and have mechanised those things in a way that was very, very difficult to do prior to the information era.

Now computers do a lot of the work of clerks, they do a lot of the work of production workers, doing repetitive activities and so on. That has led to a type of polarisation because it has had the effect of displacing labour from many middle-skilled, middle-education, middle-wage jobs in office, clerical, administrative support, production and operative positions.

When I say that that’s fine, what I mean by that is a lot of that has occurred at this point and there are many, many fewer workers who are doing simple, repetitive information processing tasks any longer. To the degree that those occupations still exist, they’ve actually changed quite substantially. There are clerical workers nowadays for sure, but they’re not simply typing and filing; they’re basically problem solvers. They’re people who coordinate events; they’re people who deal with issues of finance, and reimbursement, and complexity.

So, the point that I’m making about the limits of polarisation is simply that much of what can be readily automated as repetitive information and tasks has been done, and the frontiers are elsewhere; the frontiers are now moving into a higher-level abstract reasoning and tasks that require some mixture of creativity, and intuition, and expertise, and also moving downward into jobs that require physical dexterity and some cognitive flexibility.

It’s not that we’ve reached the limits of how computerisation will change the work that we do; it’s just that that particular strand that has been so important has not completely played out but is not the frontier at this point.

So you reckon, basically, that the computerisation, the worst is behind us, in a sense?

No, I wouldn’t say that; I would just say, “The future is different from the present.” I don’t even say, “Worse,” because I don’t think it’s been bad, but I’m just saying that the shape of it will change. A lot of the computerisation has been most evident in the so-called ‘middle-skill activities’ and that process is to a substantial extent complete, but that means that it will move into other activities. Whether that’s worse or better depends very much on who you are.

In general it’s a good thing; it corresponds to rising productivity and ultimately that’s welfare improving, but of course it has distributional consequences as well. If you’re doing a job and all of a sudden a machine can do it cheaper, that’s almost never going to be a good thing for you. If you’re doing a job and a machine makes you more productive at doing that work, that’s almost always a good thing for you.

In general, technological change like trade is not Pareto improving. It increases the set of possibilities, it increases aggregate wealth, but it does not have the effect of making every single person better off.

Yes, we’ll come to the distribution of this in a minute, but don’t you think that a lot of other jobs that have so far been largely excluded from these kinds of computerisation processes are susceptible?  

This morning, for instance, there was a programme on the BBC where the reporter visited a law firm in London where now, basically, an algorithm is being used to do contract work, which is basically text analysis – effectively replacing, what they said, three junior solicitors. Don’t you think that in these sorts of areas there’s much more to come.

Yes, absolutely; that’s exactly what I’m talking about. Previously that type of work was what we would have called ‘abstract intensive’; it was fairly impervious to automation because it required too much… It required a great deal of cognitive flexibility and solving ill-structured problems.

Now we have more machine learning, artificial intelligence type devices that can start to do that type of work. That’s not what you would call ‘polarisation’ in the typical sense, because that’s obviously targeted at a higher segment of the labour market; so, definitely.

We’ll see the opposite as well; we’ll see increasing… We’ll see machines doing things that require the dexterity of not necessarily highly educated or highly paid service workers, but things that were relatively immune to automation because machines were not flexible enough. So, I think the example you gave exactly corroborates the point I’m making.

Okay, but where do you see the balance in terms of substitution and job creation? Basically, there seem to be three categories where technology will have an impact on work. It’s either substitution, it’s augmentation or change of the existing job description, and it’s obviously job creation, but there’s a big discussion, as you know, about the balance between these three, so where do you think the balance lies?

I think that people are extremely unduly pessimistic, bordering on hysterical, both among popular speakers, business people, and even some economists, about the likelihood of vast job destruction. There are three reasons why I say that; one is that machines complement us in many, many ways and augment us.

That has always been true and that has been the source – the technological advancements that have complemented us have been the source – of our improving quality of life, for hundreds of years. One could have made the argument 100 years ago that between electricity, and mass transportation, and telecommunications there just wouldn’t be stuff for people to do, and nothing like that has come to pass. That’s one, we neglect the complementary.

Two, we neglect the fact that as we create wealth, which we certainly do through productivity improvements, we create more consumption. People want more experiences; they want more goods and services. Consequently, as people get wealthier, they tend to consume more, so that also creates demand.

Then the third thing is that we underestimate people’s creativity, and ingenuity, and how marvellous they actually are in thinking of new things. There are so many important businesses and activities now that just were not imagined 50 years ago. We have been pretty good at keeping ourselves useful and thinking of valuable things to do.

I think that absolutely many types of work will be displaced, many types of work have been displaced, but to look at the economy and say, “47% of jobs are at risk,” that’s silly; that’s thinking like an engineer, not like an economist. The idea that if work can be done by a machine and it is done by a machine, then there’s just less work to do, that’s what we call the ‘lump of labour fallacy’ and it’s not an accurate description of how economic growth actually occurs.

So, you reckon the net balance might be… The overall amount of work might even be growing, rather than shrinking, and people will remain in work? Work might be different, but it will still be usual work?

There are many nuances to this; I don’t want to make it simplistic. For one, we may have more leisure, but we have ways of dealing with that. In fact, we have more leisure now than we did 100 years ago. We work shorter workweeks and we retire earlier relative to death. In other words, we have more years of retirement, not because we retire so much earlier but because we live so much longer. So, we have a lot more leisure than we used to, and that’s a good and a bad. It can be a challenge if people are involuntarily unemployed. On the other hand, it can work fine if their work is shared to some degree.

Take, for example, two countries: Norway and Saudi Arabia. Both of them have huge amounts of sovereign wealth. You could say it’s like they have a machine that creates wealth for them. It’s not a computer, it’s just oil, but that’s okay; it creates surplus. You could say, “In those countries, maybe no-one needs to do anything, because they just have so much money,” but Norway and Saudi Arabia have handled this completely differently.

In Saudi Arabia only a little bit more than 10% of the private-sector workforce is Saudi, and the rest is guest workers. That is a recipe for long-term economic and social problems. In Norway just about everybody works, men and women, much more than most other European countries, but they don’t work that many hours. They have kept themselves relevant, and engaged, and prosperous, and actually pretty happy if you believe the data.

So, there are ways to deal with the challenge of abundance, but it’s not a bad problem to have on the scale of social problems that one could face. That’s what we’re talking about here, it’s the problem of abundance – in other words, abundant productivity, abundance of capability to do things with machines that we used to require human labour and toil for.

There are challenges that come with that. One is the leisure challenge; the other is, of course, some skills become less relevant faster than others. The people who have clearly been affected by the thrust of technological change, over the last 30 years especially, have been low-educated adults whose skills are more closely replaceable by automation, not actually necessary in the lowest-skilled jobs, but many have been displaced from middle-skilled jobs.

If I’m a clerical worker or a production worker and that type of work no longer exists, I can still do table waiting, I can still do security, I can still do cleaning, and actually I’ll probably displace an even less-educated worker who wanted that job. For example, in the US we see very few teenagers anymore working at so-called ‘teen’ jobs; they’re held by adults.

It does create challenges and they are distributional challenges. They’re not ones of not enough work to do; they’re challenges of not everybody’s skills are valuable or scarce, and so they may not be good jobs and they may not pay well. That’s the situation, so the challenge, in my view, is not about running out of jobs; it’s about the distribution of income and the inequality of opportunity. I think that is a much greater challenge.

You’re touching on two very important points. Some of the people I’ve talked to, and also in the kind of research and publishing I’ve done myself on the subject, part of the polarisation is exactly what you described. For instance, just assume that a truck driver becomes unemployed because the trucks become self-driving. The new job that person might be able to find is probably a lower-value job; it would be as a waiter.

Absolutely.

Therefore, you have a polarisation where either you move socially downwards and a few might be able to move upwards, but the threat seems to be, for most people who get displaced in the short term, that you can’t obviously train them to be a software engineer overnight, so the likelihood that they’re moving downwards on the social ladder is higher than moving upwards.  

The second point that you mention as well – and that’s obviously one of the key problems – is the distribution of the gains from this technological dividend. That is obviously against the backdrop of quite unequal societies we’re already having. So the question is, against the backdrop of these societies we already live in, how will this distributional question be solved?

You mentioned Norway, for instance; Norway has a nice way of dealing with it through its sovereign wealth fund. Some of the arguments we’ve been discussing were also whether capital ownership, given that the dividend is most likely to go towards the people owning these robots and the technology, whether capital ownership should be rethought as well, along the lines either of the public as such, via governments, via sovereign wealth funds, via different investment vehicles owning more of the capital that is reaping the dividends in the future, or maybe even workers themselves. That was what Richard Freeman was arguing.

Sure.

What kind of ideas do you have in this area, how do we rethink this distribution question?

I think these are excellent questions. I’m not ready to declare the current system – income distribution based on capital ownership plus labour scarcity – I’m not ready to declare that dead. I still think we are very much in the realm where we can tweak without having to rip it out root and branch.

What I mean by that is first of all lots of people are making lots of money from labour, labour income, and most of the growth of inequality is due to labour income inequality, not capital income inequality. That story about “It’s all about ownership of the robots,” the data don’t support that idea. I know it’s intellectually appealing and it’s exciting, but nothing suggests that that’s really what’s going on. It’s really about labour income; that’s where most of the growth of inequality has occurred.

For example, Thomas Piketty’s book Capital in the Twenty-First Century, it doesn’t even apply to the US; there hasn’t been much growth of capital inequality, at least by the accounts of the best US data available.

We have ways of moderating the consequences of inequality that have some scope still to work, so there are two… There are really three things; the first is investment – investing in people’s skills. Education is the best long-term tool we have for fighting inequality, by allowing people to be successful by selling their skills, their labour, through the labour market.

That’s how most of us make our incomes: we have skills, we acquire them, we build them, and then we sell them over the course of our career, to add value for our employers or for our customers or whatever. That’s the best way, but of course you can’t do that overnight; you can’t change the educational composition of a population. To someone who’s lost their job at job at midcareer, you can’t simply say: “Oh, well, guess you should have studied something different.”

The second tool we have is taxation, which gives the government resources to address these problems. The third, of course, is we have a system of transfers and supports that can on one hand mitigate the risks that people face through health insurance, or unemployment insurance, or disability insurance, through basic income programmes like food stamps, for example, we have in the US.

The other is that we have transfer programmes that subsidise or support work. In the US, the leading example here is Earned Income Tax Credit, which basically is a wage subsidy to workers who have low earnings potential. That both increases earnings and increases employment; it’s some mix of the two. In fact, we don’t know what exactly the mixture is, unfortunately.

Those programmes can work reasonably well, but they work much, much better if you have a reasonably well-educated population. If you have a lot of people, for example, who are illiterate and innumerate, there’s almost no amount of wage subsidy that’s going to make employers want to hire them in an advanced economy, unless they’re doing maybe agricultural work, for example, or some very low, mostly manual labour.

These policies are complements, not substitutes. It’s easier to use taxes and transfers to ensure the quality of life and maintain employment for low-educated workers if you don’t have too many of them and they aren’t too low skilled.

One problem that we face in the US, more than in Western Europe, is we have really let our educational system fall behind since the 1980s, or really since the mid-1970s, where we’ve had very slow rates of college education growth, especially among men. Our employment problems in the US are concentrated among men. Among women education has risen a lot, earnings have risen a lot, and employment rates rose a lot up until the 2000s.

Something else the US needs to do is get behind some of the family supportive policies that many European countries have done that have made it easier, for both men and women, to work as well as raising families. I hope that wasn’t too much of a digression there.

No, obviously everybody has highlighted the significance of education policy; that’s the best equipment you can basically provide to your population. The question is whether this is a sufficient or just a necessary condition for dealing with the future.

We don’t know, and anyone who tells you for sure that this time is different has a degree of certainty that is, I think, unwarranted. Anyone who tells you that this time can’t be different is equally expressing certainty that they shouldn’t, they don’t have any right to have.

Every time is somewhat different. There’s no question that technologies that we’re adopting now are different from the ones that brought us so much wealth and prosperity through the 19^th and 20^th centuries, and they have different consequences. They affect different jobs, and they will cause different types of dislocations and different types of growth.

Those who say, “There’s just no way we’ll have enough employment” are basically making a very bold claim, which is that people… They’re basically saying, in two sentences, they’re saying, “One, I can’t imagine what people will do, and two I’m certain that no-one else can either.” Therefore, nothing will happen, because no-one has the creativity to think of new things to do.

Yes, I think what pretty much everybody agrees on is that we simply don’t know what is going to happen, that we have different versions of educated guesses.

I think, actually, where I disagree, I think there’s a lot of arrogance of prediction, a lot of arrogant prediction we won’t solve this problem, that we will not think of things for people to do. That’s certainly what you get from Martin Ford’s book The Rise of the Robots. That’s what you get from Frey and Osborne about the 47% of jobs that will be displaced. It’s basically a very bold prediction on the failure of human ingenuity and creativity to think of new things for people to do, and I would never make such a bet against humanity.

On the other hand, it’s also inaccurate to say that “It’s never been a problem before; therefore, it won’t be a problem this time.” Technological change has always been disruptive, it’s always created winners and losers, and this time could be worse or better than other times.

Let me say I’m much less pessimistic than many. I’ll tell you one reason, actually, that I think is underestimated or underemphasised in this discussion is that the rate of change matters as well. It’s not just where we’re going, it’s how fast we get there, because we can only adapt so rapidly.

If we knew, if we read today in The Guardian or The Wall Street Journal that 15 years from now no-one will be driving vehicles anymore because they’ll all be done by machinery, you’d say, “That’s good, but it’s going to create some challenges. We’d better stop training people to be lorry drivers and get them ready for other occupations,” but we could deal with that.

If it was announced that coming next Monday no-one will be driving vehicles, that would be a much bigger problem – not that it wouldn’t have the same economic benefits of safer, cheaper transportation, but we would have a lot of displaced workers to contend with.

It matters how fast things are changing, not just where they’re eventually going to go, and the evidence is not strong that they are changing extremely rapidly, in fact. The productivity statistics don’t show it, the investment statistics don’t show it. I think there’s a lot of enthusiasm and certainly there is no question that progress is occurring, but the sort of singularity thinking that we’re approaching this singularity – you can see it where just the rate of change is accelerating; it’s all Moore’s Law and stuff – that’s just not serious. There’s no serious data that support it.

One of the less well-covered aspects, I think, in this discussion is that even though you can say from the technological side Moore’s Law applies and things like that, it doesn’t necessarily mean that the velocity of change is actually equivalent.

Not at all.

Because there are ethical, legal, power obstacles of actually implementing this.

Also, you have to distinguish between qualitative and quantitative change. My computer can run Microsoft Word 1,000 times faster than my computer could 20 years ago, but it doesn’t make it 1,000 times more productive; maybe it’s 20% more productive. The point is there’s this false equivalence drawn between computing processor cycles and productivity or output, and it’s really diminishing marginal returns.

To give you an example of this, I was at a conference and an executive from McKinsey got up and said, “Your washing machine today has more processing power than the entire Apollo moon project.” He meant this to demonstrate the great rate of change and the fantastic progress, and to me that just said, “Diminishing marginal returns.” My washing machine is not going to the moon.

It’s still just washing the laundry, yes.

Exactly, it just says, “When processing power is practically free at the margin, we use it for things that have very little value at the margin.” No amount of washing machine power is getting us to the moon.

The point I’m making is many of these quantitative increases, these reductions in the cost of computing, they’re great, but at the margin many of them have very, very little value. That’s what they’re going to be used for; those are not qualitative breakthroughs or enormous productivity enhancements.

Towards the end of this discussion, let me put to you another argument that goes in a similar direction that is at the moment being discussed here in the UK under the title ‘Post Capitalism’, by Paul Mason. Paul basically argues that given that, as you said, obviously economies are based on scarcity, the pricing mechanism is based on supply and demand, which is based on scarcity, but if more and more services and products are based on information, and information is abundant, the typical pricing mechanism simply doesn’t work anymore.

What Paul then predicts is that there is, I would say a more complementary new type of economy developing alongside the capitalist economy, he would say a large part of the capitalist economy will be replaced by different types of transactions because the economies are no longer based on scarcity but on abundance.

Yes, I think you see this in certain areas where information goods are very, very cheap, but I think there are many things that we consume that are in fact quite scarce – having to do with food, as an example, or experience goods like being in a beautiful location, with actual transportation and the crowdedness of roads and congestion, the cost of energy, education and healthcare. These things are actually real material goods that are quite scarce and that are costly to produce.

Information is an input into many of these things, just like you could say, “Water is not scarce in many places, so we don’t spend fortunes on water in places that are not arid.” But that doesn’t mean the economies break down, even though water is used in just about everything, because there are many, many other things that are scarce.

In general, as long as things are complementary, as long as more information then improves the quality of products that I want to purchase, then the falling price of information goods actually makes those products more valuable. Whatever inputs go into them, whether that’s silicone, or electricity, or liquid crystals, or gold, or use of the Sierra Mountains, those things become scarcer, more valuable as the complementary inputs decline in price.

My personal hunch is also that we’re going to see a much bigger variety of different economic activities.

Absolutely.

My hunch is that, as you rightly say, the long-term trend has been that what we call ‘typical working hours’ is likely to decline.

Absolutely.

Depending on the distribution of the automation dividend. We also see the breakdown of what economists used to see as the typical distinction between work and leisure, by new business models: peer-to-peer economy, sharing economy, these sorts of things.  

There will be an in-between type of economy, if you wish, and the kind of information-abundant economy that basically creates value, without any sort of monetary transaction being attached to it. My hunch is that we’re going to move into a much more colourful economic life.

Yes, I agree with you there. My real concern is I’m much more concerned about the consequence to the developing world, where a lot of the wealth of the developing world comes from two sources – three sources.

One is commodities, which I think will continue to be valuable; two is basically doing labour-intensive work for the rich world – for example, textiles and leather goods and so on, where dexterity, manual flexibility are critical and valuable because they can’t be done by machinery at present.

Then, of course, the third thing is basically domestic trade and domestic labour. There’s a lot of worker flows around the world that are basically people coming and doing domestic work, and some construction work in the rich world.

I’m concerned that automation has the potential to erode the comparative advantage of the developing world in labour-intensive goods and services. That, just like less-educated US adults are increasingly not employable in the advanced economies, I’m concerned that automation, to the degree it substitutes for labour, the labour it’s most likely to directly substitute for in the future are people doing these dexterous tasks.

That’s not a huge problem for the rich world, because we don’t have that many people doing that. If we did, it would in some sense increase our wealth, especially because we would own the technologies themselves, but for the developing world that’s a valuable resource that they supply to us. That’s a greater concern for me.

So, maybe we are looking at the wrong type of inequality if we just look at inequality within countries, rather than global inequality what it means for the international division of labour.

Yes. I would say the last 10 years, last 20 years, have been really, really good for the developing world. The question is whether that could change, and especially how many more countries can do what China has done, which is get rich through manufacturing. Will manufacturing in fact be as important when we can do more desktop printing? Will the assembly of electronics, like what China does a lot of, be important when we have workplace robots that can do more of that assembly?

As I said, I’m not worried nearly so much for the rich world; I think most of the things we do are not that automatable. We own the technology; we’re going to benefit from it vastly. The developing world, there’s more of an imminent degree of challenge.

So, that’s a whole new branch of the discussion that we’ll have to open, but we’ll have to do it at some other point, I’m afraid, David, because we are running out of time today. Thank you very much for joining this discussion.

Great.

I’m sure this wasn’t the last time that these topics surfaced.

I hope not.

Thank you very much.

This contribution is part of our project on the future of work and the digital revolution.