Unlocking Jobs in South Asia: The Potential of AI and Trade

[Deepshikha Sikarwar]
Hello and welcome everyone to this event, Unlocking Jobs in South Asia: The Potential of AI and Trade. My name is Deepshikha Sikarwar, and I would be moderating this event. We will be discussing two very topical pertinent issues when it comes to development in South Asia, Trade and AI, and what potential implications they can have on job creation and growth in the South Asian region. This event is part of the recently released World Bank Report, South Asia Development Update: Jobs, AI, and Trade. We have a very exciting lineup of speakers who will help us delve further into these issues. Before we actually begin, to all those who are watching this event online on World Bank Live, Facebook, and X, you can post about the event using #SouthAsiaDevelopment, and our live chat experts are also on the standby. You can post any questions in the live chat box. Let's begin the event with a special fireside chat. The chat is being conducted by World Bank Vice President for the South Asia Region, Johannes Zutt, and he is in conversation with Abhishek Singh, who is the CEO of India AI Mission. Over to you, John.

[Johannes Zutt]
So first of all, Abhishek, thank you very much for agreeing to have this chat with us. We really appreciate it on the World Bank Group side. Obviously, you've had a very, very distinguished career, and you are now in a very important position with respect to the approach that the government of India is taking to AI, which is a gigantic challenge, I think, for the country, and a very, very important challenge, given that business services, which are, of course, highly IT dependent, have been such a growth driver in India over the last 20 years. Helping India to make a fruitful transition to the AI-driven future is possibly the most important single thing that the economy needs to do over the next 5 to 10 years, or maybe even 5 to 10 months, at the speed that things are going. I've been in Washington for a little over two weeks now, and I was struck when I landed at Dulles airport at the number of AI processing centers that have been built just around Dulles airport. There must have been 30 or 40 of them that I could see in the plane as I was coming down to land at the airport. And that's not the only place where I've seen them. They're cropping up in all sorts of places, and they're in the news here all the time. There is a gigantic investment taking place here. And I often wonder how ready other countries are in the world for this revolution. And in particular, since I work on India and I'm the Vice President for the South Asia Region, where India, of course, is a very prominent, the prominent country, I wonder a lot about how ready India is for the AI revolution. Thank you very much for joining us today and for agreeing to share some of your thoughts. Let me just go straight into the questions if that's okay with you.

[Abhishek Singh]
Yes, sure.

[Johannes Zutt]
Okay, fantastic. As you know, our report finds that AI adoption promises a large boost to productivity in South Asian countries, especially among those 15% of South Asian workers who are in jobs where AI strongly complements human labor, such as CEOs, doctors, teachers, lawyers. Indeed, AI skills are increasingly demanded by firms with substantial growth in AI-related of job postings and a wage premium of about 30% for AI skills relative to other skills, especially in the ICT industries. Excuse me… You are the head of India AI, the organization that is responsible for implementing the government's ambitious India AI Mission. In your view, how can governments in the region help firms in South Asia take advantage of the productivity boosting potential of AI?

[Abhishek Singh]
Yes, thank you. Thank you, Johannes. It is indeed a pleasure to be talking to you and the World Bank team. What we have seen around the world, as you rightly mentioned, is that AI is impacting almost every sector of the economy, and it is also making whatever we do more efficient, more productive, saving time, being able to do more with less. I always give this example that whenever we use AI, like when I use a ChatGPT or when I use a Gemini or any other engine to do my things, what I'm able to do, I'm able to more with less. I'm able to analyze documents faster. I'm able to compare documents much more quickly. If I have to even look at balance sheets, I'm able to analyze it much, much quicker. If I have to produce a document with my own thoughts, I just need to have pointers, and then the document gets created. The non-productive work, the time taken to do the non-productive work of writing the contours of a document and wherein the core substance is there, that has become much, much more efficient and has made me more productive. So the question comes in, how can we translate these productivity gains to other sectors of the economy? For example, how do I make a farmer more efficient? And what we have seen in India is that there are use cases in which farmers can be given the right advice with regard to what fertilizers they should be using, what crop they should be sowing, what pesticides they should use, what agri-advisories, where they should sell, at what price they should sell. So ultimately, a farmer with a more informed decision because of AI advisories becomes a smarter farmer, becomes a more efficient farmer. And earlier also, this information was available on websites and apps, but very often in rural areas and in India, especially in farmers who are not very tech savvy, they found it difficult to download apps or navigate a captcha or go to on a browser and type out stuff. But if they're able to do the same thing with voice by using gen-AI tools, then they are able to quickly ask, My crop is having this element, what do I do? Then, the farmer is able to get the better advice. His input cost goes down. He is able to do more efficient operations. His income goes up. He becomes more profitable. Similarly, if you look at MSME, the smaller businesses and retailers, the shopping outlets, very often there are many of them who do not even know what product mix they should have. If they are sourcing raw materials, where they should buy from, at what price point they should buy it, how they should sell, what percentage of premium can they charge on products which are in greater demand. There are no data analytics or no AI modeling for that. Today, there are tools which are helping them become more efficient, more productive. We are seeing productivity gains across the sector. India has a large number of software engineers and IT companies. So today we see that with the tools, with AI tools, a software engineer is able to write more code in less time. He is becoming more productive. And as a result, the company also becomes more productive. We see a lot of efficiency and productivity gains across the whole ecosystem, and that is what companies are doing. That is what we are seeing, even in the government it is happening. The only flip side that remains is that when you are able to do more with less, what happens to those people who are in a way turned redundant? But that is the issue that we have to solve through scaling and reskilling. And I think that is the main issue that we need to also address while we are benefiting from the technology.

[Johannes Zutt]
Great. Thanks so much for sharing your thoughts there, Abhishek. We have been talking mostly about individuals whose productivity will be enhanced by AI, although you touched a little bit also on the other side of the equation, which is the roughly 7% of South Asian workers whose jobs are at risk from AI automation. This number is lower than in other emerging markets and developing economies. But there is a concern that AI could disrupt some segments of the labor market, especially entry-level white-collar jobs. We are actually seeing that inside the World Bank Group. Following the introduction of ChatGPT, monthly job listings for the most substitutable white-collar occupations have fallen by around 20% relative to other job postings, with the largest relative job losses having occurred in business process outsourcing and information technology. What do you think about the potential labor market impacts of AI and the best way to manage those impacts? And how is the government of India collaborating with the private sector to address AI-driven labor market transformation?

[Abhishek Singh]
As I mentioned towards the end of the earlier section, this is real. There will be some jobs that will be done better by AI bots and AI tools. And whenever we look at such job losses, the basic premise there becomes that it's not that AI is taking away jobs. People with AI skills are taking away the jobs of people without AI skills. What needs to be done is that more and more people need to acquire those skills in order to ensure that they don't become redundant. For example, if I have a software developer, if somebody is there who is writing code, if he is writing better code with GitHub Copilot or with Cursor AI, he will retain his job. If there is somebody who is still stuck in the old way of doing, he will lose. There will be a lot of need for governments to ensure reskilling and upskilling to take care of the adverse impact of this technology. Those people who are in the risk of losing jobs, they will need to acquire skills either in the field of agentic AI or the field of robotics, wherein new jobs will be created. There are a lot of jobs that will come in the data analytics space. So again, those skills will become important. What we are trying to do in India is that we have a massive program for reskilling and upskilling professionals. And also at the school level and the college level and university level, we are ensuring that any discipline anybody is studying, whether it's not necessarily STEM, science and engineering, but even students who are doing medicine or doing law or doing commerce or liberal arts, they also do a basic minimum course in the field of data science and AI so that they retain those skills in order to be relevant in the job market. So yes, the white-collar jobs will be impacted greatly, but the only way to deal with it will be through education, through scaling, reskilling at all levels.

[Johannes Zutt]
Great. I think this just reminds all of us that the world is changing and we need to change with it and that we need to ensure that when we look at our own jobs, that we remain current and acquire the skills that will enable us to remain relevant in the work that we are doing. What we've seen is that AI has a lot of potential to benefit social sectors such as education and health. I wanted to ask you a little bit about that. There is some evidence from emerging markets and developing economies that AI-powered tutoring can improve student learning outcomes, for example. But harnessing these AI benefits may require the development of specialized AI models and tailoring them to local contexts. How is India AI thinking about promoting locally suited domain-specific AI innovation within India? And how is it collaborating with the private sector on that agenda? Do you see a role for large foundational models and small language models? How do you see those differentiated? How can initiatives like AI Kosha support the creation of culturally and linguistically contextualized AI solutions in India's sectors such as health and education? And we could add also the agriculture sector, which you spoke about earlier.

[Abhishek Singh]
The points that you mentioned are like something that we are working day in and day out in order to ensure that we are able to reach the benefits of this technology to the last mile, and ultimately the end citizen should benefit. Use cases and applications become very relevant, especially in the field of education, health care, agriculture, scaling, and logistics. For example, as I mentioned, the example of agriculture, I'll take health care. There are multiple hospitals, multiple rural areas across India, wherein there are hospitals, but they would not have specialist doctors. For example, if you are diagnosed with a disease like tuberculosis or cancer or diabetic retinopathy or cataract, sometimes the number of specialist doctors required to do screening is too high and we are not able to reach the doctors at all the remote locations. But there are AI tools working on image recognition which can help in the pre-screening and ensure that we are focusing only on the really vulnerable sections of the society. Similarly, there are voice-enabled solutions wherein people can ask questions about preventive health care. During COVID, it was very, very relevant because COVID was, again, a communication challenge in many ways because people didn't know what asymptomatic meant and why they should stay indoors even if they are not sick. So again, use of AI for such communications in multiple Indian languages became very, very important. What we are doing also is supporting building such applications. And again, in education, I could give an example that the AI tutors we have, the personalized AI tutors, are proving to be much, much more effective and efficient than the professional teachers who are there in the schools. In fact, for my own children, I find that sometimes they keep on querying a ChatGPT or an AI tool in multiple ways to understand something. For example, if they have to understand the Archimedes principle or the Pythagorean theorem, the kid can go on asking in different contexts, and the AI tutor will never get bugged. In a public school what I see is that sometimes the teachers go by the learning levels of the top 5% of the class. And if there is a child who is asking stupid questions, they sometimes are shunned and the children over a period of time becomes quite shy, while asking what will the others think? And the teacher might think they are dumb or stupid, and they would stop asking questions, and that kills learning. These AI tools are very patient. They never get bugged. They keep on explaining as many times as you want. It has a huge potential, and they work in all languages. Our focus is to build both large language models and small language models so that whether it is health care, whether it is agriculture, whether it is education, we have the tutors and we have the bots. We speak the local language, which give examples of the local context. For example, if I have to explain gravity, I will explain a context to which a child can relate to. It becomes very important to train models in Indian languages with our culture, with our heritage, our context, and ensure that the data sets are fair and balanced to avoid any bias in the AI models. So that is the objective because of which we are supporting development of Indian LLMs and Indian SLMs in these sectors. In addition, what we are also doing is that ultimately for a model to become bias-free, you need data from all sources, data from all segments of society. Diverse data sets should become part of the training of the AI models. To this objective, the AI Coach platform that we have built in, which provides data sets from various domains, both public sector and private sector, and that becomes available to AI developers and deployers to train their models and develop applications and do that. That way, what happens is that while we are supporting research and providing compute and providing funding support to build models, we are also trying to build applications, especially in the key sectors like agriculture or education, health care, which meets the demands of the society, which meets the local needs and are adapted very fast.

[Johannes Zutt]
Fantastic. Well, I think that enriching the AI global experience by capturing all of the Indian thought based on a very unique culture and language, religion, history, traditions, and so on is a huge benefit that India can bring to the world. And obviously, the world is going to be looking for India to make sure that that actually happens. Thanks very much for that response. In your long career in the Indian administrative service, you have worked a lot on the use of technology for improving public administration. I want to ask you about the potential of AI for increasing the effectiveness of public service delivery and other aspects of governance in emerging markets and developing economies more generally. How do you think that AI-based solutions to public administration problems can be developed and scaled up? And how can AI be integrated into DPI-led delivery to strengthen governance and service delivery in India?

[Abhishek Singh]
Governance is one of the classic use cases of AI because a lot of jobs that we do in the bureaucracy and in the government at various levels are very often, very routine, rule-based, and can easily be done by a bot in a much, much more effective, efficient, objective, and without any bias, manner. And then it also helps in bringing efficiency in everything we do and bringing objectivity, removing subjectivity. And what we are doing in India is trying to use these solutions in multiple ways. For example, we have built-in language tools which allows us to work in multiple languages, whether it's translation of documents, transliteration, making services available in multiple languages, enabling voice-based services is something that empowers people because very often, citizens need demystifying the government. And very often, they don't understand where the information is located. Even if we are online, even if we have portals and apps, but to figure out and find out which particular link leads to where, how do you get scholarships, how do you get pensions, how do you apply for a driving license, or how do you apply for a passport. It can be a very tiring experience for somebody who is not very tech-savvy. But if the same services are made and enabled through an AI interface, wherein one can just query or give a voice query to a toll-free number or to an app and then get the reply back in your mother tongue, it can be very empowering. What we have also seen is that in judiciary cases, there are a lot of pendency of court cases. And very often what we saw, our Supreme Court looked into that matter, and they felt that a lot of time is taken on the judges and the judiciary in doing research with regard to any petition that comes up that's brought to their notice. So, we implemented a pilot wherein we have an AI tool that summarizes the new cases that come in. And then it gives a brief, neat, one-page summary and also identifies what is the point of law that is under challenge or that is under question. And then if there are similar points which have been deliberated earlier on some other case, those references are also thrown up. Then, the research time of judges is reduced. It makes them more efficient. What we also tried in urban governance is using AI for identifying issues like non-cleaning of garbage or potholes or

[unintelligible]
missing or streetlights not working. Any civic communities not functional, we have AI-enabled tools which help us in identifying them quicker and faster. Then we have data-based decision-making. We have built several models in which, suppose you want to pull together data from multiple sources and help a public service official intake and taking a decision, then what was otherwise would have required a lot of time and manual effort and sometimes lead to errors or subjectivity in decision making, that is also resolved through AI. By doing this and developing these solutions might seem to be easier, but the real challenge comes in adopting because we need to do a lot of scaling and capacity building of public service officials in enabling them to know that their jobs can be done by tools which might be as good as them in writing a note or drafting a letter. We have a huge program for capacity building and scaling of civil servants, empowering them with tools, giving them exposure, hosting them in local infrastructure so that all these concerns about data going elsewhere outside India and other servers is resolved. It is a combination of identifying use cases, building solutions, helping the departments adopt it, deploy them, and also bringing up to date, especially with regard to skills of civil servants and government officials, to be able to use them and deploy them. It is all

[unintelligible]
of several efforts that has helped us in doing this. And what we are looking at is that AI is, in a way, turbocharging the DPIs. In India, we have used technology for transforming public services in multiple ways, whether it is identity project or payments or DigiLocker or projects in health care and agriculture. Bringing a layer of AI makes it more efficient, more effective and makes it more accessible. And what we strongly believe is that many of the experience that we have of building AI-enabled DPI solutions in India will have resonance across the world. Many countries, the global south, are already looking at some of these examples. And just like DPI, we have shared as a global good to the larger community, AI-enabled solutions can also... We can work together to build up a repository of AI-enabled solutions which can benefit people at large across, not only in India, but outside India as well.

[Johannes Zutt]
Great. Those are some great examples of the way that AI can help all of us. I think it would be very hard to argue against a way of analyzing data and responding to it more rapidly. You are painting a picture of a more friction-free future where citizens' needs are understood and responded to in a more rapid and complete way. And I think that is really wonderful to hear. I want to thank you very much, Abhishek Singh, for joining me today and sharing your insights on this important topic, and also to thank you for your leadership of the India AI Mission, a very, very critical part of India's pathway to Viksit Bharat 2047. Thank you very much for being with us.

[Abhishek Singh]
Thank you. It was a pleasure talking to you. Thank you.

[Deepshikha Sikarwar]
Thank you, John. And thank you, Mr. Singh, for such insightful discussion on the impact of artificial intelligence on labor markets and smart adoption of AI, what it can do in terms of transformation, promoting innovation, growth, and boosting productivity. After this fireside chat on the impact of AI, we are moving on to the next part of this event. And we have Franziska Ohnsorge, who is the Chief Economist for South Asia for the World Bank, and who is also the lead author of the report that we are discussing today. I'm also joined by Anushka Wijesinha. He is a Sri Lankan Economist and the Director of Centre for a Smart Future. We also are joined by Amit Khandelwal, Hahn Professor of Global Affairs and Economics, at Yale University. But to begin with, Franziska, I would like to ask you, because you are the lead author of this report, on the impact of AI and the jobs, so on, so forth. But we have seen in multiple publications, focus on jobs is a very recurrent theme in most of the World Bank publications. Now, how big is this job challenge for the South Asia, particularly, if you could enlighten us with that?

[Franziska Ohnsorge]
Good morning. Good to see you. Thank you for inviting me to this. The jobs challenge is big, and it's a two-fold challenge to create more and to create better jobs for South Asia's very rapidly expanding workforce. Just to illustrate a bit the scale of the jobs challenge, since 2010, South Asia has created about 10 million jobs per year, 10 million jobs. And that compares with an increase in a working-age population of 16 million people per year. So, 10 million jobs for 16 million people every year. And that is using the UN definition of working age, it goes 15- to 64-year-old. If you use everyone from 15 years onwards, you have the working age population growing by 19 million people per year and only 10 million jobs created per year since 2010. And as a result, now the employment ratio, so the share of employed over the working-age population, is about 10 percentage points less than South Asia than in other emerging markets and developing economies. And it may fall further because of continued rapid population growth on the order of 15 to 60 million people per year. Unless job creation picks up dramatically on the order of about one-half faster, a lot of these new labor market entrants will find themselves without a job. And some will try to emigrate. As it is, South Asia lost about 0.2% per year of its working age population in 2023 and 2024. But many won't be able to unless they are modelled. So that is the model stand. And then there is the better jobs stand for those who do find a job. More than 80% of jobs in South Asia are informal, with low wages. They tend to be low wages. They tend to have very few protections. About 40% of the jobs in South Asia are in agriculture. And labor productivity, which economists use as a proxy for wage growth, labor productivity of real wages. Labor productivity in South Asia is about one-third that in other emerging markets and developing economy. And that is the reason why this need for better and more jobs is the reason why we have focused our analytical work, our research on the jobs challenge for the last two years.

[Deepshikha Sikarwar]
Franziska, we just heard John and Mr. Abhishek Singh from India AI Mission on the impact of AI while it can be transformative, but it has implications for the labor force, particularly, and we heard him say, what impacts could it have, especially on moderately educated young entry-level workers. There is always some decline we see in terms of jobs, so on, so forth. How can more jobs be created for these workers?

[Franziska Ohnsorge]
Yes, thank you for asking. Like any big technological change, AI benefits some more than others. And the two numbers to bear in mind for South Asia are 15% of workers will benefit because they are becoming more productive, and 7% of workers are at risk of displacement. Now, the incentive... Abhishek already spoke about this need for reskilling, but the incentives for individual workers to reskill are already enormous. There is a 30% wage premium. That is a strong incentive for every individual worker to reskill, even for the ones who are currently at risk of displacement, because a lot of the tasks, some of the tasks will actually easily be replaced by AI, and workers who do a lot of these tasks are at risk of displacement. And we are beginning to see that in the hiring data. We looked at job postings, millions of job postings since the introduction of ChatGPT in November 2010. For those workers who mainly do tasks that can be done by AI, job postings in these jobs have fallen by double digits behind the other job postings. We are seeing the impact of AI, the rapid spread of AI. We are seeing it in the hiring data. And we also see that the job postings for the jobs where workers… For those types of jobs that are most likely to be displaced by AI, the workers who tend to be in these jobs are younger. They have zero-to-five-year experience, they are less experienced, and they are not low-skilled. They are upper mid-skilled. They are neither very low-skilled because those tend to be manual workers, nor are they very high-skilled because those tend to benefit from AI. They are just around the BA. So take computer services, for example. The industry as a whole seems to be thriving. Exports of computer services are growing a double-digit clip. But hiring in that industry seems to have fallen double-digit behind other industries. So, these jobs traditionally held by these entry-level workers that are upper mid-skilled, those job opportunities seem to be shrinking. They are all changing. Abhishek has these wonderful examples already in the fireside chat of how the job of a farmer may not go away, it might change with AI. That motivated us to look at another new job opportunity. There are maybe new opportunities opening in manufacturing. AI is a technological change that will sweep around the world. It's already sweeping around the world, even when governments are passive. But opening opportunities in manufacturing may require government actions, a more proactive government. That is why we looked at trade reforms. We find that exactly the workers that are at risk of displacement because of this technological change, AI, would be the workers that could benefit from a manufacturing renaissance that could be triggered by lowering trade costs, in particular, tariffs, is what we looked at. Just to give you a sense of the scale of tariffs, South Asia has high tariffs, currently a 16% on average, compared to a global average of 6%. And as it happens, many of the activities that would benefit most from tariff cuts, disproportionately employ the younger and the higher-skilled workers. And that is why we are hoping to get a bit more insight on this discussion about how tariff cuts can be achieved and how they can benefit job creation in South Asia.

[Deepshikha Sikarwar]
Thank you, Franziska. I'm glad you mentioned the trade aspect of your report. Now, we are moving on to the second part of the discussion on the impact implication that the trade can bring to South Asia. I would like to begin with Anushka Wijesinha. Anushka, you have advised governments, international organizations, and the private sector on growth strategies. South Asia's manufacturing sector, like Franziska just mentioned, has struggled partly because of high tariffs on intermediate inputs that has implication for the production costs. From Sri Lanka's perspective, what tariff or regulatory reforms could help attract more investments into manufacturing and ideally into sectors that will create jobs?

[Anushka Wijesinha]
Thank you. It's a pleasure to join all of you, and congratulations to Franziska and the World Bank team for putting this very timely report together. Sri Lanka was actually an early liberalizer in the late 1970s, was one of the earlier adopters of reducing trade costs. Then we had another round of liberalization reforms in the early 1990s, but since then, neglected the trade reform agenda. As the report also shows, in those periods in which Sri Lanka did those tariff reforms, we actually saw a bump in trade to GDP, exports to GDP. In fact, in the second wave of reforms in the early 1990s, there was a bigger boost than even the first round because they were coupled with other complementary reforms, particularly reforms around FDI and trade facilitation and other regulatory improvements. I think that was one important message, both from some of the evidence contained in the report, but also from our lived experience in Sri Lanka. The importance of, yes, doing those tariff reforms, but also those other complementary trade and competitiveness reforms. The other point, I think, is clearly in South Asia, as the report shows, but also in Sri Lanka, doing those trade reforms have helped Sri Lanka plug into manufacturing-oriented sectors, into global value chains, and has created jobs. However, more recently, because there has been a tendency towards using not just tariffs, but rather these odd other border taxes that we call paratariffs in Sri Lanka, those have crept in for various reasons, sometimes because of revenue considerations from time to time to plug revenue gaps by easily using border tax as a way to do that rather than fixing direct taxation, but also sneakily because of various protectionists' request from industry. As a result, over the years, Sri Lanka has become less open to trade, and there's been a decline in trade to GDP, a decline in export to GDP, and a decline in our diversification of what we produce. That has big job creation challenges, big challenges around growth. As we saw recently with Sri Lanka, big challenges around servicing foreign debt and debt sustainability. It leaves you much more vulnerable. And perhaps unlike some of our larger South Asian neighbors that have a large domestic market, a small 22 million population country like ours cannot grow without being more trade oriented. I think, to answer your question about how to attract more investment into good jobs, manufacturing-oriented jobs, they have to necessarily be export-oriented jobs and export-oriented manufacturing. For that, it is certainly to double down on the tariff, but also the paratariff reform agenda. But not only relying on that, also doubling down on those complementary reforms that boost the capabilities of the private sector, make it actually reduce trade costs, behind the border costs, investment facilitation to bring in those export-oriented FDI. And I think this is also where we need that the report makes a good contribution to remind ourselves that this tariff, and indeed paratariff agenda, is not over. It is an unfinished agenda. And while we do focus on the new role of technology and AI, I think there are some homework that we need to do in some of our South Asian countries around doubling down on those reforms to trade, as well as around tariffs, paratariffs, but also other complementary policies that enhance trade competitiveness.

[Deepshikha Sikarwar]
Thank you. I would like to go on with you, Professor Khandelwal. We heard the Sri Lankan perspective. You are leading academic expert on economic effects of trade liberalization. You have studied India extensively in your joint work with Pinelopi Goldberg, Nina Pavcnik, and Petia B. Topalova. Across your studies on various aspects of trade reforms, what were the main long-run effects on workers and firms in India that you have seen? And how does the Indian experience really compare with that of other economies in the region, such as China?

[Amit Khandelwal]
Yes, thanks. It is a great opportunity to participate in this panel. As you mentioned, I've done some work on the Indian trade experience, particularly the major reforms in the '90s. And there's a couple of, I think, key features of that reform, which actually dovetails very well from the comments that both Franziska and Anushka have made. As most of the participants in this session probably know, India embraced a large-scale liberalization of its tariff and trade barriers starting in the early '90s. One of the things that we found when looking at the data was that there was, as you might expect, a large surge in imports coming into the country during this time. If you're a business operating during this period, you might think that that business now suddenly faces much more competition than the past, and that might cause some concern, and that might lead to some of the protectionist sentiments that Anushka was alluding to. But if you start to unpack those imports, you'll start to see that, in fact, the big surge in imports came not from what we would call final goods or consumer or durable and non-durable goods, but actually in parts and components, in intermediate components and in capital goods. That was the big source of the gains in imports. Then when you start to unpack those types of imported intermediate goods, the big gains were coming from imports from richer countries and from products that India had never actually imported in the past, what we call the extensive margin. A quick take on what that liberalization experience meant is, yes, imports grew, but it really grew substantially in terms of the amount of intermediate components and new types of intermediate components that were coming from more advanced countries, so presumably higher quality, that were now suddenly available for domestic producers. That is one key feature of the reform. Then you start to look at, well, how does that actually change the production behavior of companies in India? One of the things that we realized or that comes out from the data is that those companies start to manufacture many new types of goods for the consumer market. You see this big expansion in the number of product lines that domestic Indian companies are producing. You can pin that to the availability of new types of intermediates that are coming in. I think one of the key messages of this is that companies… International trade is quite different today than it was 30 years ago, 40 years ago. In the past, the way that a country built the industries was that you would build an industry, so the auto industry in the United States was built around Detroit. There were parts and components suppliers all within that region. Due to declining trade costs and declining communication costs, production has fragmented around the world, and now companies operate in a global value chain. In order to become a major international exporter, you need to have access to the highest quality imports, in particular, intermediates. And so two-thirds of all global trade comes from the trade of intermediate components, not final goods. I think India's experience is a good example of that. You see businesses start to reduce their production costs, their marginal costs of production fall, and that actually enables them to improve their profit margins, and they use those profit margins to invest into new product lines.

[Deepshikha Sikarwar]
So, if I could come to you, Franziska. Now, South Asia and the report also highlights. Among the EMD regions, it's the least open to international trade. Now, many would say that this protection, for instance, in form of high tariffs, protects the South Asian workers. Now, what evidence that brings you to conclude that reducing tariffs would benefit South Asian workers?

[Franziska Ohnsorge]
Thank you, Deepshikha. The region

[unintelligible]
of every country in the region, other than Maldives, is in the bottom quartile of emerging markets and developing economies. So, it is really much closer than other emerging markets in developing economies. And it is in part because of these 16% on average tariffs versus a global average of 6%. That raises the question, who is protected by these tariffs? Currently, the highest tariffs on the order of more than 20% protect parts of the labor market that are shrinking, where employment is shrinking. The highest tariffs of more than 20% protect about almost 40% of workers. But that number is shrinking and has been shrinking since 2010. In fact, the bits of the labor market that are growing are the ones that are least protected by tariffs. So, the parts of the labor market that have tariffs of 5% or less have accounted for three quarters of employment growth since 2010. Bottom line, currently, the highest tariffs are protecting a large part of the labor market, but a shrinking part of the labor market. The action, job creation is really happening in the least protected parts of the labor market. And that brings to Amit's comment about intermediate inputs. That is really important. The sectors that are currently least protected are actually hindered is really hindered by these high intermediate tariffs. Take, for example, manufacturing. Currently, the manufacturing sector in South Asia pays tariffs on intermediate inputs, imported intermediate inputs that are double of those in other emerging market peers. That is really a drag on competitiveness. If you could now cut these tariffs on intermediate inputs in particular, you could see a whole renaissance of manufacturing, and you could see a real take-off, not just in manufacturing production, but also in employment, because this is a part of the economy where employment is actually growing.

[Deepshikha Sikarwar]
Anushka, Franziska has made a compelling case for openness in South Asia. The report, the South Asia Development Update, also made this compelling case for trade openness for job creation. But the world, as we are seeing now, the geopolitical lines are totally fragmented. There are countries that once advocated lower tariffs, they are now going for higher tariffs. They are increasing, not decreasing tariffs. In this backdrop, what role do you see for South Asia, South Asian economies, to play in the international trade relations? And what can the governments in South Asia do to best support workers in the backdrop of these shifts that are taking place?

[Anushka Wijesinha]
Yes, I think interestingly, South Asia has struggled with the intra-regional trade integration compared to regions like Southeast Asia. Asia, for example, is perhaps a better place to whether some of these external big trade policy shifts because they have a lot of global production network integration within the region, and South Asia just has not seen that level of integration. I think the number that often gets cited in intra-regional trade is around 5% or less. I think the first would be to pursue greater intra-regional trade integration. Now, I think revisiting some of the tariffs and other trade barriers that we impose on each other in South Asia, sometimes can be even more

[unintelligible]
than what we have for other countries extra-regionally. That is certainly true for several countries in South Asia. I think that would be one. It's a long, unfinished agenda, but I think we have to keep plugging on at this intra-regional trade integration. That is perhaps at a policy level, but there is also a lot that can be done at an industry level, there is interest among industries or among entrepreneurs for more intra-industry trade within our region. Different countries in our region tend to have different capabilities in particular sectors. For example, in apparels, Sri Lanka may not be the most cost-effective wage arbitrage location, but we have a lot of capabilities around technology and product innovation, deep integration with buyers. What can we do to leverage on the manufacturing capabilities within our region where different countries can collaborate on different parts of the supply chain? The same is true in some of the parts and components for the automobile sector. We see the rapid growth of the Indian, particularly the South Indian automobile clusters. There is increasing interest to explore how Sri Lanka could plug into that. Several of our companies make very high-quality wire harnessers and impact sensors and rubber tires and other components. There is increasing interest to look at intra-industry trade and integration. I think that is another thing beyond the larger trade policy, trade integration agenda. A quick one on your point about supporting workers. I think this is where it goes back to some of the points mentioned earlier around training and retraining workers. I think we tend to focus quite a lot on a typical tertiary education, whereas the vocational training and TVET agenda gets left behind. I think really aligning that worker retraining to align to these new opportunities that I mentioned, plus shifting the emphasis from protecting jobs to protecting workers. We actually make it quite hard in Sri Lanka for workers to move. There could be industries that decline, that become less competitive. We are losing wage cost arbitrage-based competitiveness in some sectors, whereas there is new opportunities in others. The severance costs are extraordinarily high, so we really need to move to better ways of social protection to help workers and also help movement of these workers into different sectors, and in fact, the movement of capital as well. I think we have a lot to learn from some of India's bankruptcy law reform efforts. It is certainly a factor for Sri Lanka to consider because there will be capital that gets locked in to industries and sectors that may not have a future, but we need to make it easier for that capital to move into these new GP and linked sectors that Sri Lanka still is doing quite well in, and we'd want to double down on that.

[Deepshikha Sikarwar]
We heard the Sri Lankan perspective and what needs to be done there. Professor Khandelwal, one of the key dimensions in your research is the critical role of import of intermediate inputs. Now, why are these intermediate inputs really relevant from assessing impact of trade liberalization? And what role do you think the imported intermediate inputs have really played in the Indian economy?

[Amit Khandelwal]
As I mentioned earlier, the world is now living in a global value chain trade world. In order to produce successfully for both your domestic market and for the international market, you need access to the best quality components. The best quality components may be found inside your borders, but they are also likely to be found outside your borders. What we found is that virtually across every country in the world, the largest exporters in a country are also the largest importers. Having access to high quality intermediate components is going to improve your competitiveness, whether be it domestically or internationally. That's something that comes out of the Indian experience. It comes out of the Chinese experience, it comes out of the Sri Lankan experience, the Bangladesh experience, it comes out of the experience of Sub-Saharan African countries who were able to take advantage of exports into the US market only when they had access to high-quality intermediate components from East Asia. This is just the nature of trade today is that the production facilities have been spliced across the globe. It is just access to intermediates is just one feature that comes out over and over across studies in many different contexts as that is a key component of making a business competitive.

[Deepshikha Sikarwar]
Both, Professor Khandelwal and Anushka, they agree with what the report says on lower tariffs. Franziska, one aspect you mentioned earlier, that the younger and highly skilled workers, those in manufacturing, are the ones who would benefit the most from the trade reforms. Now, what about the less skilled people, the older workers, and those who are outside of the manufacturing sector? What are the policies that the South Asian economy should bring about to ensure that these workers can also benefit from the trade reforms?

[Franziska Ohnsorge]
Thank you for asking that, Deepshikha. That is an important question because the aggregate gains from trade may be unevenly distributed. We have all learned that. There are two policy challenges, really. One of them is one that Anushka has already mentioned. Like any other big shift, some gain more than others. And the way to have everyone gain is to have people reallocated to better, more competitive jobs, firms, places. You really need people to move, workers to move around, capital to move around. Capital is something Anushka mentioned, land to move around. Anything that can reduce the switching costs, there is this reallocation, anything that can oil the reallocation from less competitive activities to more competitive activities, will make everyone better off. For example, we have modeled in the general equilibrium model, a trade reform, and a trade reform with a labor market reform, not labor market reform per se, but a reform that reduces job switching cost by 5%, so not very much. But it just makes it easier for workers to switch jobs. And it turns out that even with this modest job switching cost reduction, you can double the gains from trade for everyone. One set of reforms, one policy priority is simply to make it easier for people, capital, land, to move from less productive to more productive activities. And that can be many things. Yes, Anushka mentioned, for example, the severance cost in Sri Lanka. So that can be labor market reform. But that can also be bankruptcy reform, something Anushka mentioned. It can also be simple plain linear connectivity, just making it easier to live in one place and work in another place. It can be skilling, because skilled workers tend to find it easier to move jobs. It can be size-dependent policies that are streamlined, so you have less of this threshold where firms don't want to grow any further because regulations start to bite. So, you encourage firms to create more jobs. It can be many things that simply make it easier for people, capital, and land to switch from less productive to more productive jobs. So that is one policy priority. And then the second policy priority is to protect the vulnerable groups. There will be some who will not be able to move. And the main thing here is, yes, the most vulnerable need to be protected. But the protection is ideally designed in a way that it doesn't lock people into unproductive jobs. I think that is something maybe Amit mentioned before, Anushka. The best social benefits or best protection of workers is the kind that gives them the option. Take, for example, a cash benefit compared to input fertilizer subsidies. A cash benefit to a farmer will allow the farmer to make the choice. Does the farmer want to move into a non-agricultural job and spend the cash in that move, the famous job switching costs? Or does the farmer want to buy fertilizer, say, with that cash benefit and continue being a farmer? At least the cash benefit gives that choice. If you just subsidize, say, fertilizers, the farmer doesn't have the choice. It is locked into farming activities. It is very important that the protection of vulnerable groups be done in a manner that doesn't lock them into unproductive activities.

[Deepshikha Sikarwar]
While we are discussing the social impact and what needs to be done in terms of measures, Professor Khandelwal, your recent research has focused on the geo-economic realignment of global supply chains. And until recently, we saw multinational corporations moving their strategies around evolving in South Asia in light of China plus one diversification. Do you see this reallocation stopping prematurely given South Asia's tariff uncertainties?

[Amit Khandelwal]
Yes. This is something that we have looked at in some earlier work in response to the first Trump administration's tariffs that were launched primarily on China. That was an episode where the US raised tariffs pretty dramatically on China and China retaliated. We'll come to the current moment in a moment. But during that experience, you might expect a country like India to actually benefit from the rising trade tensions between those two powers, because now that the Chinese market in the US was declining, maybe India would have an opportunity to grab some of that market share. What we found was that that was not, in fact, the case. Despite the fact that there was, at that time, very little changes in tariffs on Indian firms, there was no response of Indian exports into the US market. I think the reason that I leaned to as the potential explanation for that lack of response is exactly the set comments that Franziska has made, which is that when we think about, we've been talking about trade policy, and trade policy, your own internal trade policy and what you do to intermediate components are a choice that government and its citizens make. But there are also sets of terrorists that other countries impose on you, like we are seeing now in the current episode, which may be, in some sense, outside the hands of the policymakers in South Asia. A natural question to ask is, what is the best response in that situation? One response would obviously be to negotiate with your trade partners and try to come to agreements. But there are other things that you can also directly control, which is your domestic policy. And one of the things that, as Franziska was highlighting, is that domestic policy is a form of trade policy. If we go back to just this example that I keep coming to with the companies and competitiveness, if they are producing a hand towel in Bangladesh or Sri Lanka or in India and trying to export it into the United States, and that is going to come as a landed cost of $10 into the United States, and then the retailers in the US will sell it at some higher markup. But that's the price that the importer in the US is thinking about buying this hand towel from, and they're going to be looking for that opportunities around the world. That $10, what's embedded in that $10? There are intermediate tariffs that maybe the hand towel producer in India or Sri Lanka has had to pay in order to get the components in that hand towel. There is capital costs, labor costs, land costs, and costs of other kinds of materials. These are all parts of that $10. If you face now a 15% tariff in the US market, adding $1.50 on top of that $10, well, you need to be thinking about, well, what other kinds of policies as a policymaker, can I implement to reduce the cost of production to ensure that resources are, in fact, getting allocated to their most efficient uses? And so thinking about all the sets of reforms that Franziska had highlighted at a broad level are all going to be parts of an effective trade policy. Domestic policy is a form of trade policy, and that is something that policymakers in their own countries or even regional policymakers within countries have much more direct control over.

[Deepshikha Sikarwar]
We are almost coming to the end of this event. But before we close, Anushka, your closing thoughts. Very short. We don't have much time.

[Anushka Wijesinha]
Sure. I think we covered a lot. What I would recap is the importance of reminding policymakers in our part of the world, certainly, in countries like Sri Lanka, that this trade reform agenda is never done, and we can't afford to sit on our laurels based on previous efforts, doubling down on those reforms and really understanding what firms need. It is no longer just a theoretical or nice argument around why we need to reduce tariffs or why we need to reduce paratariffs. But by talking to firms, we know that they need these intermediate inputs. They need all of these complementary trade policies to become competitive. I think really highlighting that, really highlighting some of the evidence that has come out of this report to remind ourselves why this is an unfinished agenda in South Asia and why we really we need to double down if we are going to create good trade-oriented jobs.

[Deepshikha Sikarwar]
Your quick comment, Professor Khandelwal, just closing.

[Amit Khandelwal]
Yes, I agree entirely with Anushka's comments.

[Deepshikha Sikarwar]
Franziska, would you like to just wrap up the South Asia report, it emphasized on trade liberalization, the impact of AI. Would you like to just sum up as we close?

[Franziska Ohnsorge]
Thank you, Deepshikha. To me, this is a fantastic opportunity at the current juncture, it holds so much promise if policies can just seize the moment. On the one hand, you have AI creating potential for huge productivity gains, really, economy-wide, like we heard from Abhishek and Johannes. On the other hand, if policy can now get the trade costs down in whatever form, like Anushka just said, in many ways, and perhaps also break open some of these markets with new free trade agreements who really have to get the access to new markets, we could be at the start of an incredible growth burst if it all comes together.

[Deepshikha Sikarwar]
Thank you, all of you. First of all, Abhishek Singh from the India AI Mission, Johannes Zutt from the World Bank, Franziska Ohnsorge from the World Bank, Anushka Wijesinha and Professor Amit Khandelwal. Thank you, all of you who have been watching this live event on the platform. And thank you all for this very insightful discussion on the challenges that South Asia faces, South Asian economies face, and the measures that they can undertake to support the people, create the right jobs that are much needed. Thank you all, once again, for joining us.