[Indermit S. Gill] I want to welcome all of you to ABCDE. Thank you so much for coming, all of you. I want to thank Georgetown University, especially the School of Foreign Service. I want to thank Dean Joel Hellman for this partnership, and we are doing a lot, we are doing a whole lot with them. Normally speaking, we would have had the second day on campus there, but because we wanted to make it easier for all of you, we decided to have both the days here. But I especially want to thank Erwin Tiongson, my friend. Where is he? Where is Erwin? Why are you at the back? [Audience laughs]

[Indermit S. Gill]
You are responsible for this fiasco, so you should be up here. Okay? Okay. So, I want to thank him for all the work that he’s done. And by the way, I’ve been in his shoes. It’s much harder to do something like this when you are the head of a program at a university. At the World Bank, we are used to lots of resources, but I also want to thank somebody who has actually marshaled all of these resources really well, and that is Kenan Karakülah, who’s a Senior Economist in our Vice Presidency DEC. Thank you immensely for this, and they’ve prepared a feast of ideas for you. But for you to really enjoy this feast, you must bring an appetite, okay? And to bring an appetite, in this case, you must be ready to ask questions, you must be a free thinker, at least for two days. You can go back to your World Bank huddle and so on afterwards, but for the next two days, ask embarrassing questions. We have some very, very, very good people here who will be able to ask things. I mean, starting with me, right? So, here’s what I’m going to do. I’m going to tell you about findings from four reports, four of which are finished essentially. Three have been— two have been published. One has just been published a few minutes ago. And then, we also actually want to use these discussions to help shape the next World Development Report, which is on jobs for the next generation, okay? And which is being led by David McKenzie. I don’t know where David is. Here he is. And by the way, David also has a connection now to Georgetown because he’s going to actually take up a job there as soon as he finishes the WDR. That’s our deal, okay? [Audience laughs]
[Indermit S.Gill]
All right, okay, so here is my favorite— a quote from my favorite economist, and essentially it says that our real objective is not just jobs, it’s actually productive jobs. I mean, we talk about good jobs and so on, but jobs have to be productive, and that means that people can consume more because the ultimate goal of all economic activity is not investment, it’s not employment, it’s not innovation, it is consumption, okay? You do better if you consume goods, services, and leisure, and that’s the goal of this activity, right? So, here are the five reports. The Global Jobs Challenge was written last year, but we are publishing it… I think we have published it, Joe? Right. So, it’s available now on the ABCDE website. It is a great report. It’s been written by three people, Tommy Chrimes, I don’t know if he’s here. Is Tommy here? Stand up. Tommy and Kersten (Stamm), stand up. These are the two guys who did it along with Ayhan Kose, okay? [Applause]
[Indermit S. Gill]
And essentially what they do is they take a really long look at all of the statistics on jobs across the world as well as the experience of some countries in creating these jobs, countries like Singapore and Korea, and so on. Okay, and I’ll show you a few slides from this one. The second one is Business Ready 2025, which actually has a focus on what business, or what do governments have to do in terms of becoming business ready in a way that jobs are created. And it finds essentially that the countries that need to do this the most are actually the most unprepared for it. And so, there is a big policy reform agenda ahead of them and the World Bank. The next one is Women, Business and the Law. This is one of the best reports that we write every year. And Women, Business and the Law 2026 basically shows you that the gender gaps are actually greatest in economies that need to create the most jobs, and as a result of it, a lot of talent is actually kept on the outside. And this talent is not just in terms of people who can be hired for the jobs, this is talent for people who actually create these jobs, entrepreneurs. Then there are two reports that are upcoming, one of them is the Poverty, Prosperity and the Planet report, that is in draft right now, and it should be in your inboxes in a couple of weeks, and I’ll show you a few charts from that one, because there are lots of questions about the Bank’s five focus sectors, how many jobs do they create, how many people actually already work there, do the poor work in these sectors, and so on, so I’ll show you a few charts on that one. And then finally, of course, I’m not gonna show you any charts from the World Development Report 2027, because David sent me an email last night, which I didn’t read. So, I couldn’t actually translate those into some slides, but he can. I know that he and his co-authors, actually many of them who are actually going to be presenting today and tomorrow, they’ll actually be able to talk about that. But essentially, what David’s main idea is that, look, we are sort of thinking about job creation as you get somebody a job, and then you’re done, okay? But that’s not the way the world works, because people get a job and then they get another, and then they lose those jobs, they get another, and so on. So, you actually have to create an ecosphere. You have to create a system in which people can get jobs, people can get more and more rewarding jobs and move up the prosperity ladder. So that World Development Report actually gets much deeper into the productivity effects, in a very, very broad sense, it doesn’t take a static look at these things. So, all of these other reports take a more static approach. This one takes a more dynamic approach and they’ve actually… They are marshaling a lot of data on panels, on firm dynamics and so on to actually look at the dynamics of job creation. And so, he can actually tell you more about it. Okay, all right. So, this is the report that Tommy and Kersten and Ayhan have written, and essentially what this one says is that, and the data will not surprise you, that the biggest increases in young job seekers are going to be in Asia, especially South Asia, and in Africa, and the big bulge for all of the other countries or all of the other regions are in the past. For Sub-Saharan Africa, the big bulge is actually ahead, it’s happening right now. And then, the second one is that there have generally been big gaps between employment growth and output growth, and they look at that. And then, they also put all of this in a context where you say, okay, what does potential growth in Africa look like? What does potential growth in the rest of the world look like? Which means that these job seekers, this bulge of job seekers, hundreds of millions of people who will be looking for jobs, will they actually get them in the region? Will they get them outside the region or not? And they paint a pretty dire picture of this, by the way. So, here’s how they start. They start by looking at the job challenge estimates, and they have three ways of doing that. One way is the inflow of youth into the labor market, which is the one that we use a lot at the World Bank, which is the 1.2 billion number. But they also have two other methods, and I’m not gonna get too detailed into it because then I’ll be eating into Daron’s time, and I don’t want to do that. But as you can see, regardless of which method you use, Sub-Saharan Africa, and South Asia, and the Middle East, and North Africa, actually these are the three parts of the world which are going to face huge jobs challenges. Okay, then if you look at this chart, this is a very careful chart, especially the second panel on the right. If you look at that, what they look at is when did the biggest— the largest ever youth cohort happen by region? And you see that in the case of East Asia; it happened in the ‘70s and the ‘80s. In ECA, it happened in the ‘90s. In LAC, it happened in the 2000s. In MNA, it’s happening right now. And in the case of South Asia, it happened in the last decade. And in Sub-Saharan Africa, it is happening right now. So, the two regions that are actually facing the biggest cohort right now are the Middle East and North Africa and Sub-Saharan Africa. And so, it’s not a coincidence that we’ve asked Mohamad Al-Ississ and Rania Al-Mashat to actually talk about, to be here because they are from MNA. The other thing that you see on the left side, that shows that that is essentially a decomposition of that 1.2 billion number by region. And you see that it’s fairly high in the case of Sub-Saharan Africa and South Asia and MNA. Now, if you look at the countries that are the most challenged, and there you see that on the left side, you’ll see lots of countries like India, China, Nigeria, Pakistan, Indonesia, Ethiopia, Bangladesh, and Brazil and Egypt on the left. And then, on the right side, it’s like those are large economies. On the right side is the actual… The relative urgency of this, depending, means as a share of the working age population, and pretty much all of those countries are in Africa other than Western Samoa, I think. Now, they also then try to look at several indicators about which of these regions, what was the state of the region at the time that they faced this big bulge, okay? So, if you look at the first chart… Actually, so the first panel, Panel A, are those bulges, and then on the right side of that is what was the per capita GDP of the country, of the region, at the time that it faced this big bulge, and you see that Latin America as well as MNA are actually relatively well-prepared, or were relatively well-prepared in the case of LAC, but the others you sort of see that they actually face this at a time when they had fairly low per capita incomes. In the case of Sub-Saharan Africa, it’s striking that it’s very low, it’s the lowest ever. So, in that sense, you start to see the challenge there. Institutional quality, you don’t get any clear trend. It does not look like Sub-Saharan Africa has particularly low institutional quality, so that’s not where you should be looking at things. But then on the fiscal part, on the debt-to-GDP part, you again see that both South Asia as well as Sub-Saharan Africa, and to some extent, ECA, also faced a high debt-to-GDP ratio at the time that it was hit by that bulge. Then if you look at education-wise, you actually see Sub-Saharan Africa very poorly prepared. In general, though, you find that emerging markets and developing economies that have a large job challenge, also seem to have particularly low years of schooling. So, this is a very bad coincidence of events. This is not the full story though, because what makes this coincidence even worse is that this is coming at a time, in the case of Sub-Saharan Africa, it’s actually coming at a time when global growth potential is declining a lot, very, very rapidly. So, if you look at the lower charts over there, you see… No, actually, if you look at the upper charts, you see that those are potential GDP growth rates in the first decade of this millennium, the second, and then the third, and you see two things over there. One is, of course, there is a general downward trend in the case of East Asia, in the case of ECA, LAC, as well as MNA. There are two regions where you don’t see that. One is South Asia. So, in some sense, South Asia is probably in a better state to actually absorb this big youth cohort as compared with Sub-Saharan Africa, which are those bars in blue, which are not only declining, they’re also very, very low. And then, if you look at the lower panel, you see both the long-term trends in terms of investment growth are very worrying. They’ve been coming down a lot for the world as a whole. And investment, according to the work by Tommy and Kersten and others, basically they find that investment accelerations are the ones that are really needed for job growth. And they also find that in terms of this latest recovery, it is actually much, much worse in terms of investment growth as compared with previous recessions. So, the post-COVID investment boom hasn’t happened basically. So, that’s the first report and you should please take a look at that report. It’s on the web now and it’s really well-done. Here’s another really well-done report and this is a report that is produced by our Business Ready Unit, which is led by Norman (Loayza) and Valeria Perotti. I don’t know if Valeria is here. There she is. And Norman, raise your hand too. Okay. And they’ve done a wonderful job of this. This actually, at the end of this year, they will have the 2026 version, which will have all the economies, and they’re going to again do all of these numbers that I’m going to show you. They’re actually going to redo the numbers with the whole sample. But what you see over here is that in general, the economies with the youngest workforces tend to be the least business-ready. And conversely, actually, countries that are better prepared… So, I guess in a sense, I look at that and I see maybe there’s room for arbitrage here in the sense that there are some countries that are really good at creating jobs, and then there are other countries that are really good at creating people, and there might be a trade there. [Audience laughs]
[Indermit S. Gill]
And then you’ve got… And then they also find that actually, because this is not the same as, sorry, this is not the same as Doing Business, this is a much better report than that. They actually look at how well these laws are actually implemented and what are the supporting services that are provided to firms to actually conform with the laws as well. But basically, what they find is big challenges. So, here’s what I meant by the economies with the biggest job deficits are also the least business-ready. Okay, now, the third report, which is on women, work, and entrepreneurship, essentially what you see over here is that while about 3/4 of men work, or 3/4 of prime-age men work, that ratio is less than half for women, and those ratios are even lower for women’s entrepreneurship. And what they do is they actually look at this and they find that women actually are much more likely to work when legal frameworks are strong. And so, this is a choice. It’s not just a choice by women; it’s also a choice by societies. If you strengthen the legal framework, you actually see a very clear, a strong relationship between those two things. And there are lots of cool charts like this in the report. And if you haven’t looked at this report, I want to tell you, please do look at it. It’s one of the best reports we ever do, okay? Now, this is a report that we haven’t yet published, and that’s the reason why I’ve actually put the 2024 version there, but the results are from the 2026 version. And what this one looks at is how poverty reducing is a job creation strategy in a sense. And in particular, it looks at the Bank’s five sectors, which are our sectors of focus, which is agribusiness, manufacturing, healthcare, infrastructure, and energy, and tourism. Right? And it actually looks at all of these, and I’ll just give you a preview of this report. So, the first one that you see is a big drop in extreme poverty, but not for the last eight years. It’s flattened out, basically. The second one is actually much, much more encouraging that you actually find that inequality, within-country inequality has been falling a lot. So, in the year 2000, 70 countries had Gini’s higher than 40%, by 2024 last year, this number had fallen to 43%. So, like a 40% drop in the number of countries and so on. It’s still true that about essentially 20% of people still live in economies that are classified as highly unequal, but the story on inequality is actually pretty good. If you read the report, you’ll have to read these things carefully because our folks are very keen to emphasize that the world is very unequal, even if the numbers keep saying that the world is becoming much, much more equal. I call it the inequality mafia at the World Bank. Okay. [Audience laughs]
[Indermit S. Gill]
All right, the other thing that you see, and this is important, you actually find that there are a lot of people who work and who are still poor. So, you actually do find that the ratios of people, so it’s not just the case that, okay, you find a job, you’re not gonna be poor anymore. So, if, I mean, if you look at this, that’s a striking finding, and that’s one of the foundational findings for the next World Development Report, which is that it’s not just about getting somebody into a job, it’s about making sure that the jobs get more and more productive over time. And it actually looks at these numbers and it says, okay, well, 62% of all poor people are in agriculture, and 63% of the non-poor are in services. So, along the way, you actually have this transition. And the other thing, of course, is that poor people mostly work in non-wage work. So, these are the Bank’s five sectors, and essentially what you see over here is that the sectors of focus have about 43% of workers, and if you just look at two other sectors, agriculture and other non-farm work, they are about the same as that. Now, it turns out that agriculture and other non-farm work are basically where most of the poor are. And the Bank’s five focus sectors are essentially where very few poor people work. So, in a sense, the challenge over here is to grow those sectors and shrink these other sectors, at least in employment terms. So, it’s actually moving people from those sectors where a lot of the poor people are low productivity agriculture, etcetera, and move them into those other sectors. And they actually look at those things. I’m almost done. And this is how the World Bank has been doing over the last 20 years in terms of moving people out of agriculture into other things. And what you see is that you actually do see a big decrease in the amount of people who are leaving low-productivity agriculture. They’re mostly going into healthcare, some extent also agribusiness and infrastructure and energy, but not that much. Tourism and manufacturing have not been doing particularly well in terms of employment. They have been going into other non-agricultural things. So, there’s work to be done here. Lots of work to be done. But this is another chart that actually comes out there, which is, if you look at poverty, you actually see that most of it is not because of an increase in employment, it is mostly because of an increase in earnings. So, this is the productivity connection, because the way to reduce poverty is not just to get the job, but to actually make sure that these jobs are progressively more productive. Okay, and this is a chart— This is a quote that I got from actually Kenan, who did some research into this, and basically, this is from a 2017 paper, that there are around 200 million people in the world who are unemployed, classified as unemployed by the ILO, but there are 900 million people, or there were 900 million people, who are working poor. That means they have jobs, but they’re still poor. And so, this actually gives us the job that we have to do, which is for the next World Development Report and for the next couple of days over here, is to try to see, is this a definitional issue? Is it that we are classifying somebody as having a job when really, they don’t? They really have only 1/10 of a job. You know? That’s the first one. And then the second one, and I think this is important, is that we have to see a job as an important step, but it’s just the first step. It’s the first step to prosperity. And so, the jobs’ agenda is not just about sticking a job seeker into a job and then, khalas, you’re done, right? That’s— So, I use that term because of you guys, yeah. Okay, and so, essentially, it’s a time to reassess how we do labor economics, and this is the job of David McKenzie, not me, okay? But I will end over here with just one last slide. I know that I’m not supposed to introduce Daron, you are, but I thought that I’ll show you that there are two speakers at this conference who are actually on this list, this rogues gallery. This is from an Economist article in this week’s Economist. And actually, it’s so fresh that Daron has not read it. Because if he had read it, his face would be red. [laughs]
[Audience laughs]
[Indermit S. Gill]
But basically, what you see over here is it has a list of people who think that AI’s impact is gonna be big. Those are people like Erik Brynjolfsson and Chad Jones and others. And then, you have people at the bottom, Susan Athey, Simon Johnson, and Daron, who are less… who are actually much more skeptical about it. But I’m not going to get any more into this other than handing it over to Lucrecia. Thank you. [Applause]
[Lukwesa Burak]
Indermit, thank you very much. I would have been very happy for you to introduce your colleague, you know? There’s nothing nicer than that, but thank you. So, as we’ve just heard there, we’ve heard a little bit about the scale, not a little bit, we’ve heard a lot about the scale of the challenge. We’ve also heard about that big number and the deeper question that lies behind it. It’s not simply whether economies can grow, but whether growth actually translates into opportunity. Prosperity needs to be shared, the jobs created are good jobs, because for most people, development is experienced through work. We heard there about families earning a living, supporting a family, and building a future, and Gary Fields’ quote there about the working poor rings very, very true. So, our next speaker then has spent much of his career examining precisely these questions, these challenges, how institutions shape prosperity, how technology changes the nature of work, and who benefits and who doesn’t when economies transform. In 2024, he was awarded the Nobel Prize in Economic Sciences alongside Simon Johnson and James Robinson for his work on how institutions shape economic development and prosperity. And his books, including “Why Nations Fail” and “Power and Progress,” have gone on to influence policymakers, academics, and business leaders around the world. So, we can ignore that little chart that came up at the end. Today, he turns his attention then to what may be one of the most defining development questions of our age, how on earth do we create good jobs in an era increasingly shaped by AI, Artificial Intelligence? So, please join me in welcoming the Nobel laureate and MIT Institute professor and one of the world’s most influential economists, Daron Acemoglu. [Applause]

Hi. [Lukwesa Burak]
Pleasure. Thank you.

[Daron Acemoglu]
Well, it’s my distinct pleasure to be here, and I couldn’t imagine a more timely topic. And fortunately, or unfortunately, Indermit said most of the things that I wanted to say, so… Now I’ll have an easier time— Oh, sorry. I’ll have an easier time introducing the topic and also perhaps not spend as much time explaining to you why it’s so central. So, I’m going to talk about why it remains, and perhaps will become even more important to create jobs and good jobs in the age of AI. And in the age of AI, that is important because that is going to create a number of new challenges. And the central topic that I want to emphasize in the first 10 minutes of my talk is that poverty and lack of economic opportunity, some of which translates into inequality, remain still major problems around the world. And the most surefire way of tackling these problems, perhaps the only way, is to create jobs and especially good jobs, meaning high-wage jobs with some job security plus opportunities for career advancement. But with the advent of AI, that job is going to become much more complicated, if you forgive the pun. And no, I do not subscribe to the view that The Economist seems to have ascribed to me that AI is not transformative. I have frequently expressed doubt, and I will, that on its current path it will deliver amazing productivity gains. But the challenges that it will pose to jobs in the developing countries as well as in the industrialized world remain, and I will try to be explicit about those. I will also talk a little bit about preparedness for AI and global policy coordination. The first part of my talk actually very, very heavily draws on data and research from the World Bank, which has been at the forefront of emphasizing global poverty, inequality, and employment problems. Poverty is still a huge problem around the world, especially if you use not the 2%, which is the extreme poverty threshold, but the eight dollars, not 2%, two dollars a day, but the eight dollars a day… In some parts of the world, especially Sub-Saharan Africa and the Indian subcontinent, it’s 50% of the population (that) is still below this threshold. In much of the emerging world, it’s around 25%. So, a huge number, and eight dollars a day isn’t such a generous threshold. So, all of these people have incredible hunger for jobs, incredible hunger for consumption. Inequality, pre-tax inequality, especially in the advanced world, but in some of the middle-income countries is also increasing. Indeed, Indermit is right, in some of the emerging countries there has been a decline, but it’s still very high. And if you look at the World Bank’s data, the Gini index is about 40% in about half of the world. If you look at the World Income Inequality Database, it’s about 50% in about half of the world. So, there’s a huge amount of inequality. But even more jarring is that the share of the bottom 50% is incredibly low around the world, even in the emerging, even in the middle-income and high-income countries. That’s around 10%-15%. And that reflects opportunities, especially good jobs for this part of the population. As we heard, more than a billion people are going to be looking for jobs. Unfortunately, youth unemployment remains a huge problem. In parts of the world, it’s above 20%. Pretty much everywhere, youth unemployment is above 10%, and that’s going to get worse as more young people enter the job market, even if job creation didn’t become more difficult. And employment population ratios are low in many parts of the world because of female labor force participation. But in the developed world, in the industrialized world, there has been a decline in employment to population ratios because prime age men have also withdrawn from the labor force in a way that we had not seen at all 40 years ago. And then finally, in the developing world, it’s not just jobs but good jobs. So, about 50% of employment in parts of Latin America and much of sub-Saharan Africa and in the Indian subcontinent are in the informal economy. They are often very low quality, low paying, low stability, no career advancement, some of them are coercive and that’s not what the kind of jobs that we’re talking about. So, where do good jobs come from? Essentially, you need to bring three things together, as you might have thought. Demand, and that’s especially not demand for raw labor so much as demand for human skills and expertise. There have been periods in economic history where even raw labor was in high demand and wages increased, but typically it’s really human capital skills and expertise that are in short supply. Then, you do need the supply of those skills so that those jobs can be filled, and a set of institutions and market mechanisms to bring the two sides together. Episodes of high, rapid job growth, which then translate into shared prosperity, often come when all three of them work. For example, in the first half of the 20th century, there was a big employment boom and wage growth in the US economy. Where did it come from? Well, there was a demand for labor, especially new skills and expertise because the US was manufacturing new mass-produced goods and expanding production and changing the method of production as for example exemplified by Ford’s motor factories that I’m showing there. That created a lot of electrification and new machinery, but also a huge demand for new skills and expertise, which some of which was provided by training of the companies themselves, but much of it because the US had already had a boom of education at the same time, especially with the high school movement. And the market system bolstered by unions and a democratic process was particularly well suited to bring the two sides together. In many developing countries, we have typically lacked all three of those. And the situation again could get much more complicated. But when you look at, and I’m going to be quick here so that— because we’ve already covered it from Indermit’s talk, if you look at high growth episodes, for example, from East Asia, they are typically associated with high and increasing employment to population ratios. They are associated with low and declining poverty rates, and they are typically associated with stable or declining inequality, although with new technology you’ve also had periods in which rapid growth has gone hand in hand with increasing inequality. So, you see, for example, here for Singapore especially, there is a decline in the share of the bottom 10%. In slow growth economies, here I picked a bunch from South Asia, Africa, and Latin America, but the picture is different whichever ones you look at. The employment population is lower. It doesn’t typically increase. Nigeria seems to be an exception starting from a relatively low base. Poverty reduction, we’ve made some advances, but poverty remains still very high. And there is no sign of the bottom 10% income share going up, which together with slow growth means poverty remains an issue and no correction mechanism to get better. And this is drawing directly from World Bank research. Aart Kraay and others’ research at the World Bank has emphasized how many episodes of rapid poverty reduction also coincide with high growth episodes. But the question then remains is, how we’re going to create high growth in the coming decades? But I wanna also talk about one thing that’s very important and that sometimes gets obscured, which is that while a strong safety net is important, and we can all point to Scandinavian countries as a strong welfare state bolstering poverty reduction, the typical pattern from countries where we have rapid employment or rapid growth associated with shared prosperity isn’t one where redistribution ensures equality. It’s one where what happens in the market ensures a fairer division. So, here I’m drawing from a paper by Blanchet, Chancel, and Gethin, and from the World Income Database. And you can see the contrast comparing many countries. But let me reemphasize, Europe versus US. So, on the left, you see the US where the top 10% share has increased quite a lot. Bottom 10% share has fallen quite rapidly despite the fact that in the US it was not very equal even at the beginning of the sample in the 1960s. In the 1970s, sorry. And in Europe, you see a much milder increase in the top 10% share and a more or less stable bottom 10% share. Is this because of redistribution in Europe? No. If you look at the darker curves on both sides, those are the pre-tax, pre-transfer 10% bottom and top shares. So, you see that it is the behavior of what’s happening in the market that’s then mimicked to the post-tax distribution. There is some difference, of course. There is redistribution in both Europe and the United States. I’ll show you that in a little bit more detail in a second. But really, the big difference between Europe and the US, despite what Yusuf often claims here, is not redistribution. It’s really how markets work. In one place, markets have generated plenty of opportunities for people at the bottom and wages have grown. In the other, they have only grown at the top. This is even sharper when you look at the top 1%, which has skyrocketed in the US, on the left, and the bottom has fallen quite rapidly. You see the percentile changes on the right that I’m showing. And Europe and the US look very different in terms of pre-tax income distribution. So, is it really that Europe redistributes more? Well, yes and no. Europe has higher taxes and every income percentile bears those higher taxes, but overall, European tax systems are not more progressive. So, it’s not really… The difference is not in Europe being more progressive. Those higher taxes support more welfare spending and better infrastructure in Europe. Well, not everywhere, not in Germany, for example, but in much of Europe. But the US is fairly redistributive, actually, in terms of the redistributiveness of the tax the US is at the top. Transfers on the other hand are a little bit more mixed. The US is at the middle of the pack or a little towards the bottom of the pack, but not so different. And that’s why when you combine redistribution via taxes and redistribution via transfers, the US is on a par with Europe. The big difference between US inequality and European inequality is the inability of the US economy to create good jobs for lower-wage workers. And so, here is the picture in summary what the tax system is doing. The top 10% in the United States has a much bigger gap actually between its pre-tax and post-tax income than Western and Northwestern Europe and Eastern Europe. But it’s just that the pre-tax distribution is so unequal in the United States that you end up with a very unequal post-tax income distribution. So, why has shared prosperity not worked out in the United States? Well, it’s not a historical trend. In the ‘40s, late ‘40s, ‘50s, ‘60s, early ‘70s, the US was generating shared prosperity just like Western Europe. In both Western Europe and the US, you had very strong wage growth. You had very strong employment growth. And in both continents, inequality actually narrowed as the economies were growing rapidly. But then around the late 1970s, early 1980s, that picture reversed. As you can see, an amazingly jarring inequality trend set out in the US, which I’m showing on the left and the right, where the real wages of low education groups not just stagnated but actually fell for about a period of 25 years. So, those lines are all inflation adjusted. So, the fact that they’re going down from their peak in around 1980 is a very sharp decline in the real wages of low education groups. Why? Well, something as transformative as that has many causes, but one very important part of it is automation. That’s okay. Don’t worry. No worries. Thank you. On the right, I am showing a picture from a research article by myself and Pascual Restrepo, which looks at these disaggregated demographic groups distinguished by gender, education, age, and ethnicity in the United States, each one of those circles. You see that half of those are below the zero line, which corresponds to zero growth between 1980 and 2015. So, about half of the demographic groups in the United States have experienced negative or essentially stagnant real wage growth. But the horizontal axis is the one that’s most important, which is how much impacted by automation each one of these demographic groups is. Roughly speaking, it’s the share of tasks that they used to perform in 1980 that have since been automated. And for some groups, that’s like 20%. And those that have experienced that kind of automation have had very large declines in their real weekly or hourly— in this particular chart, it’s hourly wages. So, automation is a challenge for shared prosperity. That makes sense. That’s exactly what economic theory tells us, despite the fact that sometimes you hear the opposite in newspapers or some analysts that would tell you, “Oh, automation is going to create amazing economic opportunities for the poor.” Well, anything’s possible because details matter, but the general pattern is one where automation, as I’m showing there on the left, displaces workers from the tasks that they used to perform. Automation means machines, algorithms, robots taking over tasks that were previously done by humans. So, in that figure I’m showing that green area is where new automation takes over tasks, those workers are displaced. That’s not the end of the story. When that displacement happens, more efficient machines, at least according to a market mechanism, are taking over those tasks. So therefore, production of those tasks will increase. That will create demand for complementary activities. That’s why that red arrow there on the right shows demand for labor goes up from the remaining tasks. And that’s the reason why there is ambiguity. Automation could increase wages or could reduce wages depending on whether the green effect or the red effect is bigger, but it will always reduce the labor share. So therefore, productivity will always go faster than wages. And wages and employment could go up or down. Both employment and wages, they’re pretty tightly linked. So increasing inequality is often, not always, again, one of the things that you should expect from automation. This isn’t a theory. We’ve seen exactly this during the last 40 years, as I’ve indicated. It explains both the overall US trends and regional trends. And it is also what we experienced during the British Industrial Revolution. You also hear sometimes that we are— the British Industrial Revolution proves how economies create prosperity during processes of automation and big transformation. Well, that’s only half of the truth because the first 80 years of the British Industrial Revolution weren’t that happy and shiny. Wages stopped growing for about 80 years. Working conditions became longer and much more draconian. And some groups, for example, textile workers experienced tremendously large employment or wage declines and worsening of employment conditions. What I’m showing you there at the bottom figure, about two-thirds for the handloom weavers was the amount of real wage losses that they experienced. AI will accelerate automation. It is currently conceived as an automation technology, as best illustrated by the AGI loadstar, because what AGI means is that AI models will reach expertise equivalent to the very best, perhaps top 1% of workers in pretty much all activities, which means that they will take over pretty much all jobs. So that would be an amazingly large effect, and it would have 100%, no doubt, an unequalizing impact on pre-tax income distribution. Now, could we redistribute enough to make the post-tax much more equal? Perhaps. But again, there is no evidence from history to show that. And that’s why I went through some detail in the US-Europe comparison. Do not think that you can model a society like that in Scandinavia. That’s absolutely not what happened in Scandinavia. So really, this is a big threat to every country, but it’s a particularly serious threat to developing countries because of the need for job creation for an emerging youth of a very large size and remaining poverty and informal employment. The current direction of AI isn’t inevitable. We have to understand where it comes from because I’m going to suggest an explanation for how things may be a little bit less bad than what my talk so far may have implied. So, for that, we have to understand where AI’s focus on AGI, which then morphs into automation, comes from. It comes from the early computer science, especially Alan Turing and the people who spearheaded the AI field in the 1950s, such as Marvin Minsky, Herb Simon, MacArthur, etcetera, where the belief was that the machines will think exactly like human brains and can make rapid advances, and therefore once they come there, they’re gonna start replacing humans in everything. That vision has suffered many setbacks, but rather than weaken and make way for other ways of thinking about what we can do with digital technologies, has actually solidified. It’s like sort of a cognitive dissonance. But today, it’s become amazingly powerfully dominant in Silicon Valley. Essentially, it will be very hard for you to find anybody in the very top echelons of AI labs who don’t believe in this vision. What will that mean for the developing world? Well, we don’t know. There’s so much uncertainty about this: how quickly it will go, what can be automated, how those technologies will shape the global division of labor. But the IMF and ILO have come up with a number of reports and estimates. I’m just showing those from the two reports at the top. The bottom numbers, the blues, the dark blue on the right, are the automation fraction of jobs in the developing world by country or by region. These are, in my assessment, underestimates. They are underestimates for three reasons. One is I think the displacement may go faster than some of those reports assume, although probably not that much faster. Second, they ignore the indirect effect that changes in the global division of labor would do. For example, Mexican manufacturing had a big hit when American companies started introducing robots because things that were offshored to Mexico started being back onshored to the United States, reducing jobs in small auto-parts manufacturers in Mexico. So, that will likely happen to some degree around the globe. And then third, both of these reports classify a lot of jobs as potentially complemented by AI. So, that’s where I think I need to spend a little bit of time because I believe that a lot of what they do, and this is also reflected in the AI industry economists and reports, is actually mistaken. And so, why is it mistaken? Let’s take a step back and think about what it is that technology can do. The first thing is, as I mentioned, is automation. I went through that in some detail, so let me not repeat it, but it’s technology substituting machines and algorithms for tasks previously performed by humans. The second, which forms the bulk of what they call augmenting the role of AI, is what economists would call labor augmenting. Which is that you make workers more productive in the tasks that they were already performing. For example, you give a worker who is nailing nails a better hammer. That unfortunately does not create shared prosperity. And that’s where the confusion is. What labor-augmenting technological change does is that it makes you more efficient at what you do, but that also commodifies your skills because it increases the supply of what you do and unless everything else keeps up with it, it drives down the prices of your skills and expertise. Imagine what happened to cab drivers in the United— in London when GPS enabled minicabs to compete with the black cab drivers in finding all the complicated streets. That’s what’s going to happen with what the IMF and the ILO reports and many people think is augmenting technologies. That’s not going to be a path for creating true human complementary AI or shared prosperity. Capital augmenting, same, similar. What really has the potential is what I call new task creation. Meaning that you use technology not just making workers better at what they used to do, but enable them to do new things. And I’ll explain why that is very important in the next slide. But then, there is a fifth thing which is very important, I think, for the developing world, which is that technology simplifies some tasks that previously had entry barriers in terms of expertise or other things. And they enable lower-paid workers to perform them. What I and my co-authors David Autor and Simon Johnson call “pro-worker” is the new task creation. Expertise substitution is more mixed because it’s good for some workers, it may not be so good for others whose expertise now becomes easier to replicate. But for the developing world, it’s actually quite a good deal, as I’m gonna explain. So, why is it that new tasks are very different than making workers more productive in the jobs that they perform? Well, if you make workers more productive in the jobs that they perform, see, as I explained, that commodifies their labor because whatever they’re doing has now become abundant. Whereas what new tasks do is that they expand what they do. As a result, there is additional demand for their skills, not from what they were doing. In fact, what they were doing now will probably have higher prices because some of the workers that were doing just nailing nails now will divert and they will do more advanced construction work. As a result, now workers have a double gain: both new tasks that reinstate workers plus an increased demand for their services in existing tasks. That’s why AI that’s truly complementary to human, pro-worker, actually increases the labor share, the opposite of the pattern we have seen in every country over the last 450 years where labor share has been steadily going down and it increases employment and wages unambiguously. Now, expertise replacing technologies, as I said, have a displacement effect. If nurses can do what doctors can, that’s going to displace some doctors. But for nurses, it’s just like reinstatement. But here is the thing. In the developing world, you have a lot of low-paid and really reasonably high human capital workers. So, if you have these expertise-expanding technologies, that’s a big boon to the developing world as well. But here is the problem: we’re not using these technologies this way. There is no effort in my assessment, but again, AI industry insiders would disagree with this, but there is no effort to make AI more human complementary, certainly not for developing country jobs and not even in the developed world. But that’s not because of the tech-nature of technology. In fact, I argue that the pro-worker path is quite feasible and attractive because if you look into the details, the nature of human intelligence and artificial intelligence are very different. They do complementary things. Artificial intelligence is extremely good at things humans cannot do, like sifting through large amounts of data, large pattern recognition, but they don’t draw the more creative directions and judgment and interaction between physical and mental work that humans excel at. So, when two things are different, the best way for them is to be used in production in a complementary way, and that’s what pro-worker AI is. If you use AI so that you give better information to workers, you expand what they can do, you create new fields of expertise for them, that’s going to be completely transformative. But again, yes, The Economist isn’t that wrong. I am not as optimistic, not because I don’t see the possibility of AI, but I do not see the current efforts going in that direction. And by the way, when the computer industry, both in theory and in practice, went in this direction, which happened occasionally, that’s when we got the best outcomes in terms of technologies as well. So, people like Norbert Wiener, JCR Licklider, Douglas Engelbart, when in the ‘50s and ‘60s were at the forefront of advocating the then version of human complementarity, which overlaps in many, many different ways, with what I’ve just articulated, that’s when we got the early generation of technology that fueled the PC revolution, like the computer mouse, hyperlink, hypertext, menu-driven computers, and the internet. All of those came out of the labs by two people, JCR Licklider and Douglas Engelbart. That’s how transformative that agenda was. But it was very quickly abandoned. But it remains crucial both for productivity gains, and my assessment of the productivity gains that The Economist via Indermit appears very negative, which is not entirely incorrect, would be much more positive if we could use them for pro-worker ways because rather than automating, which affects just the tasks that you are transforming, and often not so well if the automation doesn’t work and doesn’t get integrated, if you make existing workers more productive, that’s completely transformative. It has much better social effects because it improves wages and employment. That’s what we needed, as we talked about. And it also creates possibilities for industrialization and sectoral transformation, which is very important for the developing world, as Indermit indicated, and the social outcomes that go with it. I think by and large, democracy is impossible when you have huge poverty and joblessness. So, that’s also very relevant. And also, we need more effort in creating new goods and services with climate change, with aging populations, with changing global balances and all of those would work much better when AI is complementary to human creativity. But there’s another challenge to AI in the AI context for the developing world, which is that the digital infrastructure, however we measure it, is unready in many developing countries. That’s both true in the private sector and the public sector. So, this is from the IMF. What you can see on the left and the right, both in terms of ICT employment and digital infrastructure, the blue, the low-income countries are very unready. But that won’t shield them from the effects of AI, even if they don’t adopt AI because of the changes in the global division of labor. So, here is my very quick summary of that, which is that as a country, you’re going to end up in one of those four boxes. Simplifying it. It’s going to depend a lot on whether AI at the frontier, meaning US and China, perhaps a little bit in Europe, goes in automation or pro-worker AI. By the way, India is making very important advances, so that could be a game changer, and I’ll come back to that in my last slide. But it also is going to matter whether you have an inadequate or adequate infrastructure. If AI goes in an automation direction and you have a good infrastructure, you will do rapid adoption. That’s the lower left corner. You’re going to get moderate adoption of AI because things are always going to be slow, especially when you go automation, which is hard to integrate. Moderate productivity growth. That’s what The Economist was complaining, probably. But low job creation because you’re automating, you’re not really using AI and technology to create new jobs. And as a result, you’re going to have negative effects on inequality and poverty. If you instead didn’t have the adequate digital infrastructure, things are no better. You won’t adopt AI as much, but you’re still going to get the same effects from the changes in the global division of labor, and you won’t get even the moderate productivity growth. If pro-worker AI were to become a reality at the frontier, then if your, again, digital infrastructure is bad, which is true for most of the developing countries right now, you won’t get any of the benefits either. So, the only hopeful place for developing countries is at the lower right-hand corner where you have pro-worker AI at the frontier and adequate digital frontier. And then, I think there is potential for high productivity growth and high job creation. And even the global changes in the— or the changes in the global division of labor wouldn’t go so much against poverty. But pro-worker AI is not easy. That’s not where the industry is going, and the industry is more and more concentrated. Today, 60% of NASDAQ is just made up of seven companies that have essentially a largely automation agenda. And worse, I think the countries that have real need for job creation and pro-worker AI have essentially no voice in the direction of AI. So, all of the countries that are in blue are those that essentially are playing no role in shaping the direction of AI. So, you see one little bright spot there is India, which has over the last five years has made more investment in AI infrastructure, although the human capital level of the Indian workforce and market integration remain problems, but I think it is indicative of perhaps what more developing countries will need to do in order to ride this wave, especially if the world frontier doesn’t— keeps on going in a direction that is not good for the developing world. So let me conclude there by saying that I think this is a very timely report that the World Bank is going to produce, and this is a fantastic conference. I’m really happy to be part of it because the good jobs, jobs imperative, and especially good jobs imperative, is truer today than ever. But the terrain has become much more adverse, and we may be underestimating the challenges that we’ll face. But that doesn’t mean the world has to be dystopian, because there are possibilities if we can redirect AI, where AI could be a complement to job creation both in the industrialized and the developing world. Thank you. [Applause]