Adair Turner - Capitalism and Robots | 'Resurrecting the Public' Lecture Series

Arjun, thank you very much. And it's a great pleasure to be here in Bangalore. I love coming to India. I do typically only come about every eight years, but I think I should do it more frequently. And it's a great pleasure to be talking here at Azim Premji University. I'm the chair of the Institute for New Economic Thinking, which has a relationship with Azim Premji, and we're very glad to have that because we're committed to looking at some of these vital issues about the economy and the limits to markets, the limits to which you can just rely on the market to solve all problems. And that's why the particular theme of this lecture series on resurrecting, bringing back a greater focus on the public, on what it is that brings societies together, what is the role of states and the public sector, I think is a crucially important one. So I'm very glad to be talking within that series. Now, whenever you're asked to give a lecture on a certain series with a particular topic, what you always tend to do is think about what you've been thinking about at the time when you got the invitation and work out how you can talk about it within the series. I'm afraid all academics tend to do that, and it's the only way to get anything actually interesting out of us because that's the thing we're thinking about at the time. And I've been thinking quite a lot recently about this issue of automation and robots: is it coming and what impact is it going to have? So what I've tried to do is think about how fast this is progressing and what it means, and then try and link it through to what that means for the role of markets and the role of the public sector. So I'm going to present a lecture in five bits. The first is a proposition that eventually we will be able to automate almost everything, and the question is when, not if. Now, the when might be quite a long time away—it might be 50 or 60 years away—but if that is the direction in which we are heading, that is going to have a pervasive impact on the shape of the global economy and the challenges which we face well before we get to that end point. So first, I'm going to say I think this is when, not if. Secondly, I'm going to talk about something which many of you, if you're generalists, will not know about but all economists know about, which is something called the Solow Paradox, where Robert Solow, a famous economist, once asked: we keep on banging on about all this automation and productivity, and why can't I see it in the productivity statistics? So that's going to be a somewhat formal and technical bit of the lecture, and you may think why on earth have I gone off on this rather technical discussion about the nature of productivity, but I hope to show you that what I'm doing there is using a device of asking that question in order to throw light on the changing shape of our economy. Then I'm going to talk about the challenges for the advanced economies, and I'm going to ask, along with someone called Tyler Cowen, 'Average is Over': is this just an inherently unequal society? I'm going to talk about challenges for emerging economies, and one of the things I'm going to say is about ending the demographic denial. I think the idea that there's a demographic dividend from a young and rapidly growing population, which is the classic thing which I would talk about at a major business conference if I were a minister of the Indian government, is a bit of denial. I broadly speaking think that countries with very rapidly growing populations have a major problem. The biggest problem is Africa, but India's a bit of an in-between between the rapid demographic slowdowns that we're seeing in East Asia and the complete failure so far to see a demographic slowdown in Africa. So I'll talk about that, and then I'll talk about limits to markets and implications for economics. Why do I think that this is the automation of everything, when not if? Well, I do think there's a reasonable place that information and communication technology is not just another technology. There is something quite exceptional about it, and these are the things which are exceptional about it. Moore's Law was set out by the founder of Intel, and he said: look, it's rather interesting in our semiconductors, we seem to be able to progress in terms of the power that we can put on a processing chip. We seem to be able to double that capacity every two years. And the amazing thing about doubling is that over two years or four years or eight years, doubling doesn't seem all that much, but two to the power of 30 is a billion. If you double over two years, after 30 years you have increased by 1 billion times. And broadly speaking, what information and communication technology is doing is that across many dimensions—processing speed, memory, bandwidth, and communications—we are tending to double what we can do every two years, which means a billion over 30 years. And it means by the time you get a century, well, it's 10 to the power of 27, which is an awful lot, I can assure you. Indeed, I'm not sure that we have one of those trillions—I don't know how many quadrillions that is. But this is extraordinary. This is why we have in our pockets, in the mobile phones you're carrying around with you, far, far more computing power than NASA had when it put a man on the moon. And it's going to go on. If that has happened over the last 30 years, it's probably going to happen in the next 30 years. And what on earth does that mean? That the thing in your pocket will have a billion times more computing power than your computer or your laptop today. That's transformational. The other thing that's transformational is software. Software is not like electromechanical machinery. It has a very simple property: once you've made one copy of a piece of software, the next billion or ten billion or hundred billion don't really cost you anything. It has a zero marginal cost of copying. And that is extraordinary because it means that if clever people can create an app, a piece of software, and it takes them some man or woman years to create it, that can then be sold to billions of people at almost no cost. I think it's also the case, and I've been reading about this recently, that these progress that people talk about in artificial intelligence—this is for real. We are getting closer and closer across many aspects of human activity to something which is reasonably called artificial intelligence. And there's a very simple property which various people have pointed out. There's a good book by a man called Nick Bostrom called 'Superintelligence', which is the moment we create artificial intelligence which is equal to human intelligence, we will then rapidly accelerate to superintelligence because we will have created an ability to accelerate up our rate of progress above that which humans can achieve. We have this extraordinary phenomenon called big data analysis, the ability to switch on software which goes out and looks at data to find patterns where you don't even have to ask it in advance to look for particular categories of patterns. It simply finds patterns in medical data, in transport data, in weather data that we didn't know existed. And we have this extraordinary phenomenon of machine learning and robotics. Before I understood this, I used to think that if you wanted to make a robot that sewed, you would have to program it. But now, with machine learning, you can show a robot a movement, and it learns to replicate that movement. That's what we call machine learning and robotics. And I think if you put all of those together, we do have a remarkable technology which I think will probably mean that eventually we can automate driving, we can automate language translation, we can automate medical surgery, we can automate almost anything you can think of. In the end, however, it will come at very different paces by different types of activity. And one of the most important things when we start moving away from the long far distance where we've automated everything to what's going to happen decade by decade is to think about which bits will automate earlier and which bits will automate somewhat later. And there's a very good piece of work being done by McKinsey Global Institute which has stepped through like a queue.

It is thinking about what would be best to switch off. And I'll just stick close to this. We've got an echo there. If you start with that breakdown of what is easy to automate in terms of capability and then translate that into types of work activity, what you clearly get in this analysis is that the easy thing to automate is predictable physical activity—the things where people just do again and again and again the same manual function. That's highly automated. But the other end, things which require deep expertise or management functions, are difficult to automate. And what the analysis says is: of all the time spent in all US occupations, how many fall into predictable physical, processing data, collecting data, etc.? And from that you can then progress to say, well, what does this mean for automation potential by occupation? Towards the high end, you get graders and sorters of agricultural products—these are just examples from a full range—and at the end you get, at the moment, psychiatrists and legislators. Now, I'm going to come back to where the legislators are really doing something so incredibly wonderful and human that they can't be replicated. I think the attitude of a lot of people towards politicians probably doesn't want to believe that they are uniquely valuable to human beings, but clearly we have to have them in human life. But we have this spectrum of how easy it is to automate away, and that then drives a potential automation by sector. Over at the top, you get certain categories of accommodation and food service which might be highly automated, manufacturing, transportation, and warehousing. And at the other, you get some education services, management, etc. Now, I don't think this is a perfect analysis, and in fact there are very particular ones that you might challenge. I suspect there's some aspects of food service which are actually less easy to automate than much of manufacturing. But the basic idea that what we have is a whole load of activities within our economy, and some of them are easy to automate and some of them are difficult to automate, and that in particular it's the predictable physical or the ones which are mechanical and predictable processing of data which is most automatable—that I think is a crucial insight. What McKinsey finally do is to say, well, we just don't know when everything will be automated. But starting today, they suggest that on the blue bit—that's a different scenario of when things could be technically automated—and they argue that if you look at all the activities in the US, 50% technically could be automated today, and that that will slowly rise to 100%. And that if you were a real technological optimist, it could happen in a big whoosh in about 15 years' time. And if you're a technological pessimist, it might not happen for 40 or 45 years. And then you've got a different dimension which says that even if it's technically possible, it will take longer for it to economically occur. And then you've got a wider range which could take us way out to 2095 before we've automated everything. But yet, 2095 or it could occur earlier. So I'm not predicting when this is going to occur, but I am convinced that if you look at the fundamental nature of the technology that we're facing, we are going to have more and more things automatable, where machines can do all these functions quite as well as any human being can. That in the end, almost all human activity will be automatable. And that within the different activities, there is a hierarchy of things which at the moment probably says a hell of a lot of repetitive physical things, the same thing done every time, are highly automated. Which now lets that is true.

Ask the question: the Solow Paradox. Well, you know what the Solow Paradox is. For some of you, it's this: we've been talking for decades about rapid productivity growth and all these computers, etc. But if you actually look at the data on productivity growth, whether it's measured in output per person or output per hour, it actually seems to have slowed down, not accelerated. So there are a group of economists out there who said, 'Stop telling me these amazing stories about how we can automate everything. Why don't I see it in the actual productivity statistics of the advanced economies?' And it was Robert Solow who said way back in 1987, 'You can see the computer age everywhere but in the productivity statistics.' He says, 'You're telling me about Moore's Law, you're telling me about software infinitely replicable, you're telling me about artificial intelligence, big data, and robotics. Why can I not see an accelerating rate of productivity growth in the GDP figures of a society?' And that question has become even more pressing over the last 10 years because actually in a lot of countries we have seen a further slowdown in the productivity growth process. So here we see UK output per hour. What should I—I seem to be cutting in and out. Should I take this off or just rely on this? Shall I take this off and then when I walk around I'll talk very, very loud because I think that's causing a bit of a problem. Good. Okay. This debate about why we don't see the productivity growth has increased even more in the last 10 years, where we've seen a slowdown of productivity growth in many advanced economies. But I'm going to suggest that the Solow Paradox is inevitable when you allow for what I call Baumol effects and the high-tech, high-touch paradox, the zero-sum paradox, and nil or low-cost benefits missing from GDP. So let me as rapidly as possible tell you what I mean by that. The reason why I call these things Baumol effects is just for the economists amongst us: there was a famous economist called William Baumol who described the first effect about 30 years ago, and that's why I've labeled it Baumol. To understand these three effects, you have to start with what I think is the mental model that most economists have in their mind when they think about productivity growth. And the easy way to think about productivity growth—and you'll find a lot of people talk about it—is moving from agriculture to factories. As we make agriculture more efficient, you have a mental model in which we start with 100 self-sufficient peasant farmers producing 100 units of food. We work out how to produce food more efficiently. We have 50 farmers producing 100 units of food, and 50 workers are able to leave the land and they produce 100 units of cars, washing machines, televisions, etc. GDP has doubled, productivity has doubled. And in the standard model which people have at the back of their mind, this is an infinitely replicable process because of that step. Step two is we double productivity on farms again. We now have 25 farmers producing 100 food, we've got 50 factory workers producing 200 cars, washing machines. 15 factory workers have gone off to produce something even more sophisticated—mobile phones—and we've also got 10 service workers working in hotels and retail, etc. And we repeat it and we repeat it and repeat it. And each time we repeat it, we take people out of some existing activity, they end up in a new activity, and in the new activity that is subject to productivity improvement because everything is susceptible to productivity improvement. So that's the standard model in which productivity improvement in one sector means productivity improvement across the economy. But I think it's worthwhile noting that you can also have something else. You can have: you start with 100 farmers produce 100 units of food. 50 farmers produce 100 units of food, and the other 50 people turn into domestic servants paid half as much, producing 50 units of value. Agricultural productivity doubles, total economy productivity increases 50%. And in a sense, it's quite interesting to ask whether this is essentially what may have happened in the first agricultural revolution, by which I mean the creation of settled agriculture in Mesopotamia and the Fertile Crescent back in the 8th millennium BC, where you have a form of agricultural revolution as people stop being pastoralists and hunter-gatherers and become settled agriculturalists. But what it enables is basically the emergence of a leisured elite, a priestly elite with lots of servants. And actually, the productivity improvement after a one-off spurt then stops because what happens in this environment is that you don't have an endlessly repeatable process. You have what economists would call an asymptotic process which will end up with one farmer producing 100 units of food and 99 servants serving him, at which point productivity has increased by 100% but it all comes to an end at that point. It's worthwhile pointing out that this process of it coming to an end could also involve some people not going off to be domestic servants but going off to be artists and priests and singers at the farmers' parties. You could have five artists, singers, entertainers, and fashion designers paid twice as much even as the farmers. But even if you have that, if those functions can't be automated, you will still eventually asymptote to static productivity growth, even if in the agriculture itself every generation you are still doubling productivity. Now, it's quite interesting to work out why you would have this Baumol effect. What will determine whether you have what I call a Baumol effect, where things keep on moving to new activities which cannot themselves be automated? It will tend to depend on at least three things. It will depend on whether in the newly emerging economic activities there is inherently automation potential or not. It will depend on the impact of the productivity increase on income distribution. If when those farmers became more productive they all equally benefited from it, then the increase in productivity might have gone into leisure time, or it might have gone into manufactured goods, but it wouldn't have gone into servants because you wouldn't have had richer farmers employing poorer farmers. So it's partly driven by the distribution of income, which depends on how we organize our asset ownership. And it depends on the consumption choices of the winners from the initial productivity increase. But if the winners from the initial productivity increase are highly skewed towards a small number of rich people, and if those small number of rich people want to spend their money on some things which are not very automatable, then you will have an end to the automation process. Now, do we see these activities at work? Could this be part of what is going on and which explains the fact that although we're seeing huge productivity growth potential in some factories, some retailing areas, some wholesale distribution, at the level of the total UK economy we're hardly seeing productivity growth at all? Well, we might, because I would like to show you 21st century technology in London. This is a delivery driver. This chap is on a bicycle with a box on his back delivering pizzas around London. And actually, I think we see in the advanced economies quite a lot of incredibly rapid productivity growth in particular sectors of the economy—less jobs in retail because it's all moving over to Amazon—and then people need income, so they find things to do. So we get an easier ability to get things delivered to our house rather than having to go and pick it up ourselves, but we don't pay much for it, and it's not highly automated, at least not immediately automatable. It's sort of the British economy looking a little bit more like the Indian economy, with this chap wandering around on a bicycle with a box on his back. And actually, if you look at US jobs growth forecasts over the next 10 years, they are set out by the US Bureau of Labor Statistics, which I think has one of the best sets of labor statistics in the world. Every two years it updates its forecasts of where it thinks jobs will come from over the next 10 years, and it's been pretty good in the past. What you see here is where it thinks those jobs will come from and what they are paid. And if you look at the top 10, the biggest growth categories by thousands of jobs created, look at what they are: personal care aides, home health aides, food preparation and serving workers, nursing assistants, cooks in restaurants. They are those things which at the moment we can't immediately automate. Often they're things where how many of them we need depends on how much richer people want to have a personal care aide or a home health aide, etc. And on average, you'll see on the right-hand side, apart from number nine, general and operational managers, and number two, registered nurses, they are paid well less than the average. All the big job category growth is coming in a set of things often to do with face-to-face service which haven't yet reached the point of automation. They tend to be low productivity, they tend to be low paid jobs. And have a look: you have to go down to number 14 before you find software application developers. They're just much less important in job creation terms than personal care aides. So I think there is a Baumol effect at work, where we rapidly automate some sections of the economy, and then it doesn't show up in the productivity statistics because the people freed up have got to do something, so they find a relatively low paid, low productivity job to do. Are we seeing it in India? Well, I was struck by this quote from an Economist article two weeks ago about a factory somewhere in India. What was going on was the manager was explaining what will happen when he opens the crates with the new bits of factory automation equipment. He says, 'His job will go, and his over there, and that one too.' But the manager then insists that, as in the past, he'll find jobs for everyone as drivers or as watchmen. So I suspect in India we probably are seeing rapid automation, for instance, of some factory jobs, but because people have to find jobs, somebody finds them a job, but it's a relatively low paid, low productivity job. And so here's the first of the paradoxes: the more rapidly technological progress enables the rapid and radical automation of existing activities, the more we will see what I call high-touch jobs grow in activities which at least for now cannot be automated, but they may be ones where wages have to be low enough to make automation currently uneconomic. And I think that is partly what is going on in this world of radical automation of some functions in the economy not showing up in the statistics. But then there's a second thing which could be going on. Because if I go back to my standard model with my 100 farmers, suppose this happens: 50 farmers produce 100 units of food, 25 people go off to be criminals, and 25 are employed as police to defend the farmers against the criminals. In this environment, total measured productivity has increased less than before—it's only increased 25%—but there's been no human welfare benefit of increased consumption because I think we can all agree that criminal services is not exactly something that we like consuming, or police services we don't like consuming. Police do a very good job, but they're only there because there are criminals. It's a zero-sum game where these things cancel each other out. And if you get a proliferation of zero-sum activities, there are two questions that you have to ask. One: we've got technological progress which increases productivity in the agricultural sector—more efficient farmers—and we produce more criminals and more police. Does total measured productivity increase? The answer is it depends upon the conventions by which we measure GDP. And one of the things I'm going to say is we think GDP means something, and increasingly we realize that GDP just means what GDP means, and it depends on all sorts of pretty arbitrary accounting conventions. We count police services in GDP because public sector output is measured by the cost of input because we can't work out any other way to do it. It's just a convention. It's the way we do it. We could have chosen not to do it. Indeed, in the early days when people were trying to work out how to do GDP back in the 1930s and 40s, some people wanted to exclude things like police services. They said, 'GDP is meant to measure increases in human welfare. What's that to do with police services?' But essentially, there are a set of conventions. But even if total measured productivity in this case is increased, you've got a second question: does rising GDP per capita deliver increased human welfare? And in this case, there is no increase in human welfare although there's an increase in GDP. Now, I think what is quite surprising, at least in advanced economies, I think we've got a hell of a lot of zero-sum activities. I mean, I am the director of two financial services companies in the insurance space, and what I have received recently is a lot of briefing from our cyber experts about how we're going to defend ourselves against cyber attacks. And these cyber experts are really, really clever people, but they're not increasing human welfare because all they're doing is making sure that we are not degraded in our end consumer service by cyber attacks. And if you think about cyber criminals and the large numbers of highly-skilled cyber experts within companies, if you think about companies doing bad selling practices and financial regulators telling them not to do it and compliance officers then trying to work out how to make sure that they don't do it in future, and compensation lawyers advertising for the business of bringing a suit against the financial service company—that's a zero-sum activity. And pretty much all legal services are zero-sum activities. I mean, think about divorce lawyers. Suppose over a five-year period divorce lawyers got much cleverer and much better at doing their job as divorce lawyers. Human welfare wouldn't be any better because there'd be a better divorce lawyer on both sides, and it would simply cancel each other out. Much financial trading and complex financial innovation I think is essentially a zero-sum game of people competing to win versus another but has no necessarily human benefit. Servicing the purchase and sale of goods that already exist, some education services—dare I say—I think are fundamentally about enabling people to signal by the fact that they got into University X that they are high quality, to make it easy for the company to know they're high quality, rather than necessarily increasing human capital. And as for politics, elections, lobby groups, think tanks, and even academic economists—I know that's a shocking idea to agree—what are we doing? I mean, we're debating. Politics is a very important process, but it's basically a debate about how to divide up an economic cake, it's not about creating more economic cake. And let's even imagine fashion design and intensive branding activities. Fashion design creating beautiful things—yes, it's an expression of human creativity which contributes to human welfare both in the sense of the people doing it doing a creative activity and in the sense that I think it's better that we have this than don't have it. I think the world would be a drab place if we all wandered around in 1950s issue Mao suits and none of us were allowed different colored scarves or handbags or anything like that. But I think we have no reason whatsoever to believe that the fashions and brands of 2050 are going to make us any happier than the fashions or brands of 2017. This is simply a cyclical process. It's better to have it than not to have it, but it's ultimately a form of zero-sum activities. And if you get these zero-sum activities, there is then a completely arbitrary, as I pointed out earlier, question of what they do in terms of those two questions: do they come into GDP or not? Divorce lawyers are in GDP because it's household consumption, so richer, more expensive divorce lawyers, they're in GDP and in the increase in productivity statistics, even if you don't think they're necessarily increasing human welfare. Corporate lawyers and cyber experts are what we call an intermediate product within GDP, and they don't show up in GDP. So if you get lots more of them, then you'll get a slowdown in GDP growth. So you can have a phenomenon where if we automate away the whole of manufacturing and the whole of warehousing and the whole of transport and the whole of everything that we really need to do, the jobs which deliver the things which give us a reasonable standard of living, and the surplus people all go off to be corporate lawyers and cyber experts fighting cyber criminals, you will get a slowdown in measured productivity growth even though underneath it there's a sort of productivity miracle going on.

This has then some also some odd paradoxes to it. And the paradox is that actually some of the things that you might have thought are least essential for human welfare may be the ones that proliferate in this environment. If I go back to this chart, I think we will probably automate away all sorts of absolutely essential activities for human welfare, but in the short term we won't automate away legislators. Indeed, the more productivity we have, the more money that the American political system may spend on political campaigns. And there's an interesting paradox here: the more that things are completely just games, the more that they are clearly zero-sum activities in ultimate human welfare, the less they're probably threatened in terms of their income. Think about this paradox: computers can easily beat grandmasters at chess. Indeed, we're now at the stage where a computer that I could put on your laptop will beat a grandmaster even if the computer has to think in one second and the grandmaster gets an hour between each move. But this hasn't made any difference whatsoever to the pay of chess grandmasters, which is going up. Whereas when we automate away things that we really need, like manufacturing, the job and the wage rate may go down. And if in 2050 robots can beat Man United, I don't think that is going to mean the top soccer player earnings will fall. We have what I call automation and the zero-sum paradox: rapid technological progress could eventually automate away almost all of the activities which are truly essential for human welfare, while supporting increased intensity of zero-sum competition for relative income and status, so that zero-sum activities account for an increasing percentage of income and measured output over time, and where a lot of the highest incomes are focused on zero-sum activity. Are we seeing this in the figures? Again, go back to the US Bureau of Labor Statistics figures. I think we probably are. I think within not leisure and recreation but within some of the educational areas, with some in particular of finance and business services, I think we are seeing a proliferation and increase in the importance of zero-sum activities. There is something going on in our economies, and I think it's going to go on, which I call 'finding things to do'. John Maynard Keynes, back in 1929, in a brilliant short essay—it's only eight pages long, and I advise anybody to read it—it's called 'Economic Possibilities for Our Grandchildren'. He said, 'Let's imagine a world a hundred years hence'—that's not very far away now, that's 2029—and he said, 'By then we'll be able to produce all that we need for a good standard of living with a 15-hour week.' Well, we don't have 15-hour weeks in the advanced economies. Here's my hypothesis: if people had a high leisure preference and if the distribution of income enabled everyone to enjoy a good standard of living while working 15 hours a week—which might require much more equal distribution of income than we have—I think we would in the advanced economies be producing the vast majority of all the goods and services essential for our standard of living with far fewer work hours, and we would then be sitting on top of a far faster rate of growth of productivity than we actually observe. One of the things that makes me think about this every now and then: we have in Britain some sort of transport strike or an incredibly small amount of snow which makes it very, very difficult for British trains to run because when we get the wrong sort of snow, we can't run our trains. And in that, a lot of people don't turn up to work. And it's always struck me that almost nothing happens in terms of people's standard of living. And they've actually to work out how much work and how many people are required to turn up for work for us to get everything that we actually need for our standard of living—it's quite a small percentage of the total. But we find things to do. And we find things to do because we have status consideration, we have political goods, we compete for status, we need to compete for relative status and relative income because whether you live in the nice part of town is not determined by your absolute income, it's determined by your relative income. So you work extra hours to be able to do that. We have individuals who have adequate income minimum income requirements, and 15 hours isn't enough, so they find things to do, and people find things for them to do. And we have work as a social activity. People don't have an absolute sure preference. But my hypothesis is that what's been going on in the advanced economies is that underneath the productivity figures, the essential work that we've been doing is much less than the work we are doing, and that productivity increase is far faster than our measures suggest, and that that is the answer to the Solow Paradox. Now, what I've suggested so far may feel like problems. We have this rapid rate of productivity growth, and rather than us all enjoying a higher standard of living while working 15 hours a day, we're all beavering away and finding various forms of zero-sum competition to do, or we're finding low productivity marginal service jobs for people to do before they need the income. It may seem that these are problems, but my third point of what may be wrong with the productivity figures is that there may be all sorts of great benefits for human welfare that don't show up in productivity figures at all. Imagine that over the next 50 years we create a wonder drug. This wonder drug gets rid of all diseases. I mean, basically it means that we all live till 100 and then we drop dead. We don't have a long period of ill health, we don't have Alzheimer's, nobody dies of tragic early deaths which are terrible for the families and a loss of opportunity. We all live long healthy lives and then that's it, we just drop dead. I think everybody would admit that this would be a huge increase in human welfare. Where would it show up in GDP? Well, the answer is it depends. It depends how you run the drug development system, it depends how good you are at working out your price indices in terms of nominal GDP. With a system of private development and patent protection, if over time you think about what will go into nominal GDP during the research and development phase, it will be entirely dependent on the accounting treatment as to whether the research and development expenditure is capitalized or not. If it's not capitalized, it won't be in GDP. If it is capitalized, it'll count as an investment in GDP. It will then come into GDP in the period as quite a lot in the period where it's a high-priced drug under patent protection, and then the moment it comes off patent protection and the price collapses, it'll collapse out of GDP. If it's produced in manufacturing plants which get ever more efficient and have lower cost, it will eventually disappear to almost nothing. And by the end of the century, it will play a trivial role in GDP, even though looking back we'll say it's the single most important thing which has improved people's lives. It will be different if the government or charitable development develops this drug. It'll never then show up as a bigger figure in GDP because you'll never charge as much. It'll go straight from the research and development stage to the generic manufacturing phase, and this thing of equal value to human welfare will be much smaller in GDP. That's in nominal GDP. Will it show up in real GDP, in our measures of inflation-adjusted GDP? It depends entirely on how effective the statisticians are at spotting that fall in the price at the end of the period. If they spot that price at the end of the period, they'll produce a period where they say inflation is incredibly low this year because the price of these things has collapsed. But if this is our pattern, that price collapse will never occur, so it won't be able to stay in real GDP. So we live in a world—this may sound rather techie—in which all sorts of things which are hugely valuable for human welfare just don't end up in this measure which we call real GDP. The wonder drug in real GDP will, as I said, depend on the effectiveness with which national income accounts capture price reduction effects. It will only get there if you go through a period of very high price patent protection on route. Now, some economists argue that this means that we have very significantly underestimated productivity and real income growth because it isn't just new drugs, it's mobile phones and tablets, streamed films and music, computer games, and social networks. All many of these are free, all they cost a tiny proportion of what they would have 30 years ago. So it may be that what's really going on in this Solow Paradox is a rapid productivity improvement which just doesn't show up in our figures because we have falling prices and increasing quality, and we just don't look at it in real GDP and productivity growth. And Marty Feldstein therefore argues that the US underestimates growth, and he therefore underestimates that all this talk in the US about real incomes not growing is nonsense. He says the result is that the increase in real incomes is underestimated, and that the common concern about what appears to be low growth of average household incomes is misplaced. These low growth estimates fail to reflect the innovations in everything from health care to Internet services to video entertainment which have made life better during those years. Do we just have a measurement problem, that a whole load of the most important things for human welfare just aren't in our measures? Well, one caveat on that, and I think it's fairly clear that low productivity growth estimates fail to reflect super rapid productivity growth, falling prices, increasing quality, and innovation. I think in the sense of that first step, is there something going on technologically that you don't see in the productivity figures? Answer: probably yes. Does this mean that human welfare improvement has been understated? Answer: not absolutely certain. I think if this is about drugs, then yes. But actually, there's a lot of evidence in advanced economies that ever more sophisticated computer games and ever more social networks and ever more mobile phones and the whole mechanisms of how adolescents operate today is making the average adolescent in the advanced economies more neurotic, less happy, more anxious than they were 30 or 40 years ago. So remember, you always have to say: are we mismeasuring? Are we failing to see the productivity improvement in some physical sense? You can answer yes, we're failing to see it, but that's not quite the same as has human life got better and we failed to see it. Put all these three factors together, and I think Solow's Paradox is inevitable. I think we're seeing Baumol effects and we're seeing this high-tech, high-touch paradox. We've got rapid technological progress in some sectors of the economy, but we've got a proliferation of these delivery drivers. I think we've got a zero-sum paradox, and that as we get richer, as we automate away the things we really need to do, we find things to do: we get divorced and we employ divorce lawyers to fight against one another. And I think we have nil or low-cost benefits missing from GDP. But I don't think it's as simple as to whether all of those are increasing human welfare. And so I think what's going on is that our standard assumption in economics that as technological advance drives productivity improvement in every sector of the economy, which shows up in GDP measures of output per hour worked and per capita, which provides a good measure of improvements of human welfare, was never wholly true, but it was sort of adequate in the early stages of the farm to factory transition, but is breaking down in an information intensive world. And that the second step, that that provides a good measure of improvements of human welfare, is also not a bad assumption when your GDP goes from a thousand dollars a year to ten thousand or twenty thousand dollars per year, but beyond that it just isn't true. There's no necessary connection there. What is the end point of all this? And I think the end point in 2100 is that solar-powered robots guided by artificial intelligence systems do almost all the work needed to deliver welfare-enhancing goods and services. But all of this really useful work accounts for an incredibly small proportion of measured GDP, while almost all human activity is devoted to zero-sum competitive activities which account for most of GDP. And that the growth or not of GDP per capita measured on current accounting conventions by the end of this process tells us almost nothing about human welfare. There's an interesting question for economists. You non-economists don't worry yourselves about the next question, but you understand why economists will have to worry about it: does the economy even exist? Because if the economy even exists, we can't be employed as economists, and that would be fatal. In common economics, we are told in undergraduate economics that economics is about the allocation of scarce resources in consumption and production. Well, if robots can do all the work, is there any scarcity? My hypothesis is that at the end of this process, the income measures of GDP will be dominated by real property ownership values and rents. Some people will own real property and others will rent from them. Intellectual property rents: in the course of this century, we will create a group of people who are clever enough to create this new piece of software or this new song or this brand, and that will be providing a rent flow which is derived by intellectual property, by patent protection, or by the power of a subjective brand. And we will have very high incomes for a very small number of people who are skilled or lucky in information technology, in artistic performance, in the ability to create subjective value, or in processes of zero-sum competition. And it is the distribution of this monetary income which will determine the distribution and consumption of scarce positional or status goods, even if the good news is that the absolute necessary things for human life might be relatively free. And I think we see signs that we're heading in that direction already. Many of you in this room may have heard of Thomas Piketty's 'Capital in the Twenty-First Century'. Some of you may have bought Thomas Piketty's book, and a somewhat smaller minority will have read it from cover to cover because it is about 700 pages long. But in it, Piketty points out this remarkable point: that in the advanced economies, the ratio of wealth to income, having fallen in the mid-twentieth century, has soared back to six or seven times. But what Piketty himself has not stressed, which is there in the blue bit of this chart, is that almost all of that increase in the wealth to income ratio is explained by the growth of the value of land, the value of housing and the land on which it sits in locationally desirable cities. And so we have another paradox: the more rapid the progress in information and communications technologies, the more that value will be placed on inherently physical assets and attributes—desirable land, sporting capability, physical beauty. Supermodels are paid a much bigger multiple of average earnings than they were 30 years ago. So our various Bollywood stars create subjective values. Increasingly, the very process of having this extraordinary information technology which automates away everything else means that this is where the value in the economy goes to.

Okay, let me talk quickly now: challenges for advanced economies, challenges for emerging economies, and limits to markets. Challenges for our economies will be heavily driven by where the automation is going to occur. And I think we are going to see in the advanced economies the continued rapid automation of some functions: accommodation and food services. The UK Retail Consortium believes that the three million jobs in British retail could come down to about 2.1 million jobs within easily within the next 10 years, as we use automatic checkout, etc. Logistics and warehousing: we're seeing warehouses now are just automatic picking machines. Manufacturing, obviously. What does this mean? Does it mean there's a lack of jobs? No, I don't think it does mean there's a lack of jobs. I think there are two very bad arguments for saying the jobs will always be there, and these are the classic business defense arguments where I say, 'Well, my company is creating more jobs.' That's a bad argument because what you want to know is what is the total effect of jobs. But the good argument is there's no absolute limit to the number of jobs which can be created in flexible labor markets. The question is what will be the pay rate of those jobs. The fundamental issue is not the capacity to find more things for people to do, but the incomes that they will earn. And across the advanced economies, we are seeing these very, very big increases in inequality, with the top 1% pulling way away from the middle of the income distribution, while the bottom 20% does very badly indeed. And that is being driven, I think, by the power of information and communications technology at both the top and the bottom end of the income distribution. At the top end of the income distribution, what we are seeing in the modern world is extraordinary value creation with very few jobs. Facebook is worth getting on for 400 billion dollars when I last looked at it, and employed 15,000 people—that may be up to 20,000 now—but the point about it is it's a drop in the ocean of the global labor market. And it is always going to be what we have in this environment because software is infinitely replicable at zero cost. It's an extraordinary ability for very small amounts of people to create a bit of software, to create a bit of an app, to create a brand, and for that to give enormous value with very few jobs. That creates enormous inequality at the top end of the income distribution, while at the bottom we are finding things to do in low paid face-to-face service jobs. So what does this mean? It seems to have happened before actually. I'll go quickly over this. If you look at the United Kingdom from 1780 onwards, in the hundred years after we began the Industrial Revolution, for the first 70 or 80 years, real wages and the share of real labor income fell. And by 1840, capitalism was not working for ordinary people, and we came pretty close to having social revolution. And we only didn't have social revolution because we mowed people down at the Peterloo Massacre. We used the power of the ruling class to suppress the working class. But capitalism does not always—capitalism in the long term may produce a rise in real wages for everybody, but the long term can be a hundred years. More skills is an answer? I think it's an inadequate answer. I think however much you skill people up, only a small number of highly skilled people are required to drive rapid productivity growth in automatable sectors, and incomes will be set by relative skills, not by average skills. However many people are good enough coders to create a new computer game, the revenues from computer games are going to accrue to the small number of people who in any one year have made the computer game which has captured the customers' imaginations. There isn't much return in the world from being the 23rd best computer game maker in the world. So I'm passionately in favor of skills and education, but I don't think it's an adequate answer here. Nil and low-cost benefits come to the rescue? You could argue that even the low income earners will enjoy a huge flow of nil or low-cost benefits from drug development, low cost computing and communications power, nil and low-cost internet services. But we still have the problem that not all free services necessarily deliver welfare or life satisfaction benefits. Are social networks actually making us happier? And welfare and life satisfaction may be heavily driven by relative position and access to positional goods. Essentially, the low income earner in that future economy is described living in a poor location, and in order to get an adequate income to pay the rent, having to go on an enormously long commute. They may not say, 'Oh, it's all okay because I've got a computer game which is a thousand times better than it was 40 years ago.' Those nil cost benefits don't make up for the relative disadvantage in these things. So I think there are major issues in how developed economies manage the way through this. There's an interesting point put forward by a man called Tyler Cowen. Cowen is a rather interesting American economist, a deliberate intellectual profile cutter, and he has described essentially in a book called 'Average is Over' what is an essential dystopia. And he doesn't wholly believe in it, but he sort of puts it forward to provoke us. He says, 'Look, we're not going to need a whole load of people to do all these jobs because they'll all be done by machines. But we can make it work because the poorer people will live in tiny houses in middle America where there's a warm climate, where land cost is low, and they'll have adequate health care and close to zero cost entertainment. And as a result, they won't turn up in New York and lead a riot against the earnings of the top 1%. And as for the more talented Bohemians who however don't want to be part of the New York rat race, they'll all go off and they'll have coffee shops in warehouses in Detroit or Berlin where there are old cities which have low land costs, and there we have hipsters, and it'll all be great fun and they'll just opt out of the rat race.' And he says we really shouldn't worry about that rising income and wealth inequality leading to revolutionary revolt. The long-term picture will be fairly calm and indeed downright orderly. It's a deliberate provocation, and I don't know what the answers are. I don't want that to be the answer. And by the way, I don't think the answer works at all in a more densely populated place. It's easier to imagine that in a place like America where there's a lot of physical space, and you can just have people living in trailer parks with free flows of entertainment. It's much less easy to imagine that in South East England. But there are very major issues for the developed economies. How about the developing? Well, I think there's a major challenge here. The automation potential: it may be a century before all the jobs are automatable, but we start with predictable physical, which is what goes on in factories, and we start with predictable data processing, which is what goes on in call centers. So that's two immediate challenges for job creation here. And this is a major challenge because in economic history, there is no example of rapid economic catch-up which does not involve a large role for low-cost labor-intensive manufacturing. That's how Korea did it, that's how Taiwan did it, that's how Hong Kong did it, that's how China did it. But the current reality is that we are seeing what Dani Rodrik has called premature industrialization: countries where the manufacturing share of employment is falling, not like in Britain once we're already rich, but in countries which were still way, way back down in terms of income per capita. There's an ILO report—International Labour Office—which suggests that 16 to 90% of low paid jobs in Asian textiles and clothing could disappear within the foreseeable future through automation. At the Adidas Speed Factory in Germany, and SMAC in Germany, which is about to open, they have a super automated factory for making shoes. And 160 workers—just 160 workers—will produce 500,000 shoes per annum. I can't get the precise figures, but as best I can work out, Adidas' total supply chain across the world might be about a million people. If this is the productivity at which they're going to work in future, 96,000 people can produce the 300 million shoes which Adidas produces. That's 1/10 as many workers. And by the way, that will happen not just because it will be lower cost, but also it's just a much better way of serving your market. Because if that Adidas factory is in Ansbach in Frankfurt, and a whole load of people start liking that particular brand of shoe that week, you'll be able to software reprogram the Ansbach factory so that you produce more of that type of shoe and less of another type of shoe instantaneously, rather than down a complicated multi-country supply chain. So I think there are major issues here, and I think these issues are biggest for the economies with rapidly growing working populations. I hold the counter-conventional wisdom point of view that I wouldn't worry at all about China with its shrinking working population. I think that shrinking working age population will force them to robotize, to have rising real wages, and I think they'll manage to do it. I'm absolutely terrified about Africa. I simply cannot see how major African countries facing the sort of increases in working age population forecast throughout this century are going to find a route to full employment. I think India is an interesting intermediate phase. There are some very major issues about Indian growth and job creation. India is now achieving reasonably rapid growth, but it still cannot absorb the 10 to 12 million people required per annum—jobs created per annum—just to keep the employment rate stable. The employment rate is falling. Formal employment in leading sectors—the ones everybody in India always talks about: IT, BPO, generic pharma—they're great jobs, they're highly skilled jobs, but they're very small relative to the working age population out there that has to be employed. The textile policy targets 10 million new jobs, but there's a new report out by Tex Ravel and Ernst & Young that says that even if the market grows by 40%, with levels of automation, it pencils in 2.9 million. And interestingly, in India we're seeing an increasing per capita income divergence by region, by city. The top three large states back in 1980 had incomes 1.5 times the bottom three large states—that's excluding some of the very smaller states to get an overall picture—and that's three times higher. So is India going to go direct from a sort of agricultural underemployment equilibrium to a sort of form of Tyler Cowen's dystopia, with lots of fantastic leading high-tech businesses including in manufacturing, but large numbers of low paid, low productivity Baumol-type jobs, with huge income, wealth, and real estate price divergence between leading states and cities and laggards, but with at least everybody doing okay because drugs will be cheap enough that they live adequately healthy lives, and entertainment costs nothing, and clothing is pretty cheap as well? I noticed in the latest version of the India Economic Survey the interesting phenomenon that while there's a very large divergence going on in income per capita between the states, in terms of health outcome there's a convergence going on in that life expectancy is going up actually fastest in some of the most low income per capita states. And that may continue to be the case if we can create wonder drugs, the marginal cost of producing which is very low. Well, I don't think we'll agree that that's a reasonable dystopia to aim for, but it does mean there are some big policy issues and priorities in developing economies. One has really got to focus on where the job creation is going to come from. One has to maximize manufacturing success, but probably recognize this is not going to be transformative in total numbers of jobs. You've got to identify and support inherently less vulnerable employment. Where is that? I don't know. Maybe it's in construction, maybe it's in tourism. But I think that's one of the biggest issues. You've got to accept large numbers of Baumol-type jobs while ensuring adequate income. What does that mean? Does it mean universal basic income? Do we have to accept that we have to design cities, urban and transport infrastructure for maximum inclusion, to allow as many people as possible in the cities to live well there even if they don't have monetary incomes that enable them to afford the very expensive housing in the middle of the city? And we've got to achieve a true demographic dividend, which for countries like Africa means they better achieve falling fertility very fast or they've got a major problem. India is now on the path to falling fertility even in places like Uttar Pradesh and Bihar. And of course, there are many states now at or below replacement. The challenge is simply the demographic impetus from the high fertility rates 30 years ago, and you can't do much about that. But there are other countries throughout Africa who've got to begin that demographic transition.

Okay, final point: what does this mean for limits to markets and for economics? Here's the good news. I think that free market competition, supported yes by basic government research and development expenditure that drives the absolutely fundamentals of great basic science—which is something private enterprise itself will never pay for—but the combination is highly likely to drive rapid underlying progress of the productivity frontier. I think this machine that we have in the world driven by the free market is going to produce autonomous self-driving trucks, self-driving cars, better and better drugs, more powerful mobile phones. This is unstoppable, and the free market will deliver it. But the complexities and problems are that markets won't resolve increasingly important distributional issues because this world of technological cornucopia could be a world of increasing inequality. Nor will they necessarily ensure that increasing productivity necessarily translates into increasing human welfare. I think we have to think about a way in which rising income, its impact on greater utility and happiness, is of a different shape. If rising income delivers good health, that is definitively good for human welfare. If rising income means we all spend more money on fashion goods—well, I want to have that, I don't want everybody to have to wear the same suit—but don't kid ourselves it's going to make us ever happier. And if it delivers congestion and environmental damage, then it's bad for human welfare. And that implies some real issues for the role of the public space as against markets. What this economy suggests, I think, is that economics has to move from the predominant focus. The predominant focus has been on what's called the neoclassical production function. We believe that we should think about how the combination of labor and capital in a competitive market economy will lead to increased productivity. I think the increase in productivity is the least of our problems, and that the really interesting issues for economics are: what determines real property values and rents? How does that interface with urban geography? What should be the socially acceptable taxation of real property? How do we make sure that this is not radically unequal? Environmental and congestion activities. Intellectual property rents: the direction of all American policy for the last 30 years has been to increase intellectual property protection, including by extending the protection of Mickey Mouse for another ten years and then another ten years, so that the rent flows continue to flow to Disney. Well, you can't justify that in terms of incentives because Walt Disney is dead and he created Mickey Mouse already, so he doesn't need an incentive to create Mickey Mouse. And actually, a hell of a lot of our economy is now about intellectual property rents, and there's a very major issue about how long we ought to have patent protection to create some incentive for people to create artistic and intellectual things, but you can have far too much. Positional goods and financial instability, very important development challenges, job creation, income distribution, universal basic income, good public good provision. In this environment, GDP will be of decreasing importance. Funnily enough, Lionel Robbins, who wrote back in the 1930s—and I came across this because at the LSE about five years ago I gave the Lionel Robbins Memorial Lectures—in an essay on 'The Nature and Significance of Economic Science' back in 1932, he said: 'Both the concept of money income and of national money income'—by which he means GDP—'have strict significance only for monetary policy.' What he was arguing is that measures of GDP are useful month by month, quarter by quarter, because they help the central bank work out whether things are accelerating or slowing down, and you manage the economy to make sure that you don't get unnecessarily recessions which produce unemployment. But he argued even then that if you use them to try and work out whether France was better off than Britain, you were pretty much wasting your time. I think we weren't completely wasting our time then because I think actually they work better than they do now, but we are suffering from a gradual degradation of the value of this dominant measure. And then of course we have to come to our long-term challenge. Now, the long-term challenge for all economists is that whenever they think they've thought out something original, they find that John Maynard Keynes said it 70 years ago. And the long-term challenge which John Maynard Keynes set out at the end of 'Economic Possibilities for Our Grandchildren': he said that when we have this complete cornucopia of a fully automated economy, for the first time since his creation—we must apologize on behalf of John Maynard Keynes for the slight lack of gender neutrality in this statement, but let us assume that every time he says 'his' he meant 'his or her'—thus for the first time since his creation, man will be faced with his real, his permanent problem: how to use his freedom from pressing economic cares, how to occupy the leisure which science and compound interest will have won for him, to live wisely and agreeably and well. And I think what is interesting is that that is going to be quite as big a challenge as meeting the productivity challenge, which I think information technology has already met on a massive scale. Thank you very much.

Right, I think you know, see that one. I'm going to have to say I am within the next couple of weeks going to produce a full text, and I hope that the full text will answer some of the probably 15 questions that all spoke in that. I mean, just very, very quickly to one or two of the clarifications. Positional goods and status goods: it's actually a rather important distinction. A status good is something which is valued because it signals status. This goes back to the writings of Thorstein Veblen and the idea of conspicuous consumption as being something that the rich do to illustrate that they're successful. And there is an element in the incredibly expensive Gucci handbag, etc., which is a status thing. And there is an element to which we deliberately through brand, through advertising, encourage people that they've got to have that brand, way beyond what logically you would have thought that is going to do to their happiness. The point about status competition, though, is that if you engage in status competition, what matters to success in status competition is not your absolute income, it's your relative income, because you've got to be richer than the other people in order to afford these things, because the people who produce these things deliberately don't produce enough of them for everybody to afford them. They deliberately create scarcity so that they can have scarcity value. Well, the good news in relation to a status good is that if you're a person who doesn't care about that, this doesn't affect you at all because your consumption preferences lie somewhere else entirely. The fact that somebody else is desperately keen to spend a large amount of money on a status good doesn't affect you. A positional good is different. The most obvious example of a positional good is housing. It is objectively nicer when you go on a holiday to a beach to be at a hotel on the beach rather than at a hotel four miles away from the beach. It is objectively nicer when you go skiing to be able to stay at a hotel close to the skiing piste rather than down the mountain where you have to get a bus up. It is objectively nicer to live in a part of town where the crime rate is lower and the pollution is less and it's closer to where you work. And so these are things which it's much more difficult for you to say, 'Well, no, I couldn't care about that,' because they're sort of inherent. But they are also things where relative income matters because your ability to afford the hotel on the beach rather than four miles away doesn't depend on your absolute income, it depends upon your income relative to other people, because there's only a certain amount of those slots and they're rationed out by the price system. So positional goods are things the distribution of which will be driven by relative income, but they're pretty difficult for you to opt out of, or you have to be more radical in your choice of a different lifestyle to opt out of it. Status goods are simpler: things that we're competing on because they are signals of our success—that particular flashy SUV, that handbag, that brand of whiskey, even if I couldn't tell the difference in a blind taste test but I think I should, etc. So that's the difference. I think the others I'll have to leave for you to read the document. On the first one, estate duties, inheritance tax: well, within the tax package which Donald Trump is now putting forward, I think there is a proposed abolition of the inheritance tax in the US, the estate tax. And I think that's a ludicrous idea. Look, I think in almost any state of the world that we work in which there is reasonable freedom of human behavior, you will inevitably have a significant amount of inheritance. You'll never be able to stop that. If you try to have a 100% inheritance tax above a certain level, they can make lifetime gifts. Even if you can't make lifetime gifts, you can give people advantages through education, etc. So let's not kid ourselves we're going to get around that. But to actually make the problem of extreme inherited inequality worse by actually abolishing inheritance taxes is just the wrong thing to do. And my overall attitude: I'm not a socialist, I'm not proposing total equality. I think there is a value in human individuality which will always express itself in significant differences in income. I think there is some of the creativity which is required in our economy that is appropriately rewarded by higher incomes for some than others. So I'm absolutely not arguing that we can run our society in anything like a totally equal fashion. But I think what we face at the moment, and have faced for several years, is a set of free-market tendencies to increase the level of inequality in societies. And I think in that environment, it is crazy simultaneously to dismantle progressive redistribution. Indeed, I would somewhat lean the other way. And I don't know how much I believe the other way, so I'm a very frustrating person for ideologues of either the Marxist left or the neoliberal right because I don't give absolute answers. I believe in a sort of pragmatic, common-sense response to where the world is heading and trying to balance it a bit. But that's my answer. I think this question was about massage services or even sex services as the extreme form of face-to-face services. Well, I think that raises some moral issues which I'm going to not wade into. I think if I can pick another area of what I call face-to-face service: social care, the care of older people, the care of poorer people, is an interesting one in automation terms. Because actually it involves some functions which we should be quite happy to automate. People who work with elderly people in social care often have to do some really hard physical things—lifting old people up to bathe them, cutting toenails, dealing with toilet events, etc. If we can more automate that or provide physical equipment that does that, I think it would be perfectly good to take that out. I don't think there's a particular human benefit in that being face to face. But what we do need with older people or frail people, people who have Alzheimer's, etc., is people who are willing to be with them and talk to them and play games with them. And if in societies where that's not done enough by children, there are great jobs to be done. And I think we undervalue those jobs and we underpay those jobs. And I think we should be finding ways in societies which, once they've gone through the demographic challenge, will have an increasing number of old people, some of whom, before we have created my wonder drug, will have Alzheimer's, will have depression. I think we should be valuing those sort of what I call face-to-face services more. But I think that involves states which are strong enough and willing enough to impose taxation on the better off to be able to afford good quality services of that nature.

Well, that's a very nice cue for me to say that I've written a book called 'Between Debt and the Devil', and what I found was that actually there is a lack of attention in much recent literature to the role which debt plays. And I think debt plays a fundamental role in instability in modern economies. The classic trade cycle, the inventory trade cycle—the build-up of stocks and down—is much less powerful than it used to be and will become less and less important with IT and better stock management. And most instabilities in modern advanced economies are now caused by surges of lending, primarily against real estate, which then produce a crisis which produces a recession. That's what happened in Japan in 1990, that's what happened in Scandinavia in 1992, and we see it again and again. So it's hugely important from an instability point of view. It's also a very important issue in terms of inequality. And the reason why a series of religions—Islam, medieval Christianity, I don't actually know the Hindu attitude to it—and also a series of secular philosophies, it's in Aristotle as well, have been very worried about debt contracts. One of the things that debt contracts can do is take relatively small initial inequalities and magnify them. There are two farmers: one has a good harvest, one has a bad harvest. The richer farmer lends money to the poorer farmer. Play compound interest forward ten generations, and this guy's a feudal lord and this guy's a bonded serf. Debt has played a major role in that. Now, as it happens, I'm a visiting professor at the School of Islamic Finance in Malaysia because when I began to write my book on debt, various people from the Islamic finance area asked me to come and talk to them about it. And before I came, I said, 'Let me be first of all clear: I'm a completely fully paid-up non-religious secularist, and secondly, I don't believe there should be no debt in our economies. I believe there should be some debt. But I do think we need to pay much more attention to how much debt and what forms of debt.' I think we should be very careful of heavily advertised, high interest consumer debt because I think it can pull lower-income people into a position which is destructive of their economics. So incredibly important issue. Gross national happiness: look, what I'm saying is we cannot rely on GDP or even GDP per capita as being a good and total measure of human welfare. I'm a little bit wary of replacing it with another statistical shot at gross national happiness because gross national happiness is going to be another somewhat arbitrary measure in which you take a number of things and you weight them in some algorithm and say, 'There's gross national happiness.' And I'm not against doing it as an alternative measure, but I don't think we should have any measure. And I think this is actually the essence of what politics is and what the public is. We should never believe that we can reduce the process of social and political debate to a measure. In the same way that a company—not a great company, but some companies—can say, 'My one objective is profit,' societies are not like that. We can't say, 'This is the thing and we're maximizing that.' What the public space should be is an endless, never-resolved debate between alternative objectives and alternative priorities. And I think one of the worries that we have is that that idea of the public space as being that arena for that endless, well-informed debate of informed and empowered citizens is under threat. And if I can come back to the first question, I think it's under threat from fake news, I think it's under threat from a 24-hour news cycle which produces a short-termism, I think it's under threat from both right and left. I think in America, the left wing has been guilty of a retreat into pointless, self-referential identity politics which has gone down some rabbit holes of relatively unimportant issues, while the right has told a simplistic story of a wonderful market in which everything works and within which individuals just maximize their own unencumbered limits. I'm really confident that what I want—which is a sort of sensible middle of informed citizens trying to navigate this space to an economy which will be a market economy with all the dynamism and choice of individual freedom, but nevertheless modulated by state making interventions informed by a political process—well, I'm probably somewhat more pessimistic than I have been in the past. In the UK, I think we are caught between a Conservative Party which I think has taken a necessary consolidation of the fiscal stance far too far and has pushed down our tax revenues and public expenditure to a level below that which is required in an advanced society to have really good health care, etc., and we're caught between that and a Labour Party run by a leadership some of whom believe that Venezuela is an attractive model. And I think anybody who thinks that is an attractive model is living in cloud cuckoo land. So I think the challenge is there. That's why it is so important that we try and create a space for intelligent discussion to address these issues thoughtfully and try and address these challenges. And I think we should end on that note.