 Good morning. I would like to share some thoughts with you about the role of urbanization in poverty reduction. This joint work with Joachim de Weert and Yasuyuki Todo from the University of Tokyo. The world is urbanizing rapidly. It may not come as a surprise to you, but we know this. This is being talked about, but already half of the world's people live in urban areas today. By 2020, we expect the same for the people in the developing country. Half of the developing country's people will be living in urban areas. We hear about it, we talk about it, but I think what's much less known is that the urban world is also concentrating rapidly. It's not just that more people live in urban areas, but actually more of the urban people are also living in fewer cities. When I talk about concentrating, basically you talk about the number of urban population, the amount of urban population which live in one or two cities. Basically, we expect another one billion people in the developing world to urbanize, to move to cities. Out of this one billion, we expect, or the projections are, that about two-thirds, 660 million, will actually be going to cities larger than one million. This is quite different from in the past. In the past, basically over 1970 to 2010, there were about 2.2 billion people going into urban areas in the developing world. But at that time, there was only one billion actually going to big cities, to cities larger than one million. So urbanization is happening, it's happening rapidly, but more importantly, or that's what we would like to draw attention to, is that on top of that, the urban population is also concentrating. This is also happening in Africa. Already, about 40% of Africa's urban population lives in big cities, cities of one million or more. About the other 40%, they basically live in small towns and you sort of see the distribution here is from the census data, about 40 countries in Africa, 1990 to 2010. Basically, you sort of get these two bumps here, where sort of you see concentration in the big cities, more than one million, and then sort of another concentration here, you sort of center around 100,000. So this is happening and again, if you look at the growth rates over the past 20 years, you'll see that the big cities have been growing at 6.5% sort of sort of metropolitanization happening, while the small towns were growing at 2.4%. The urbanization is happening, on top of it you get urban concentration. Why should we care? One reason why we make care is to sort of read and get some ideas from Professor Henderson, Vernon Henderson who is an urban economist, a neuron article in 2003 in the Journal of Economic Growth. And basically, looking at the experiences across the world, he was finding that the degree of urbanization, whether you're 50%, 60%, 70% urbanized, as such, does not matter for how fast you grow. But there seemed to be an optimal degree of urban concentration. The amount of people, percentage of people, the share of the urban population who lives in one or two cities. So that seemed to matter for growth. There were some points at which you could be too concentrated. Instead of having a spread out urbanization pattern, too many people live in Paris, too many people or all of your urban population lives in Kinshasa. That may not be a good thing for growth at some point. Now what about poverty? What about the distributional effects? There's a comfort about inclusive growth. What about the effect of an urban concentration on poverty, the rate of poverty reduction, shared prosperity? And there's very little literature on that. It's actually both theoretical as well as empirical. There's very little to find. And yet, this is the question that countries are facing today. China, India, they're bracing themselves for mega city development. Yet another wave of urban migration in China and where the Chinese government increasingly are focusing on trying to house these people in big cities. Vietnam is facing the same challenges. Rural migrants are coming into the cities. Are they all going to be concentrated in Hanoi, Danang, Ho Chi Minh? Or is it better to sort of have a buffer refuge of secondary towns around it to catch that wave of new rural migrants? Africa, we know urban concentration is already high. And yet, again, as you put yourself into a certain position, as you put the infrastructure in place in these cities, in doing that, you also lock yourself in a certain pattern. So if there is some relationship, if there is a relationship between the rate of poverty reduction and the degree of urban concentration, then basically you're setting yourself up for a certain pattern of development. So that's why we would argue that this is important. Now, again, it's not so obvious immediately why it would matter or how it would matter. What could be reasons behind this that sort of a metropolitanization would lead to slower or faster poverty reduction versus a rural town more spread out urbanization pattern. So if you look at the literature, one reason it could be sort of has to do with agglomeration economies. If people come together, live together or densely together, then basically you get activities are concentrated and there could be economies of scale. On top of these economies of scale, if your supply is closed by, if your customer is closed by, the service provider is closed by, you sort of get all kinds of spillover effects that reduces the transaction cost. Or even further, there could be knowledge spillovers. So all these, this very fundamental, new in 15 years, it's high on the agenda. The idea of agglomeration economies and how agglomeration economies drive economic growth, sort of intermediate you have an urbanization process in there, would sort of make an argument that from that perspective, it may well be big cities, it may well be metropoles, which are important for economic growth, does job creation, and then depending on who gets the jobs also poverty reduction. But it's sort of from an agglomeration economy point of view, there's a big argument for sort of a metropolitization. Now in this literature, even though it's a very dominant and very strong argument, there are also quite a few caveats in terms of the empirics of it. These agglomeration economies may not be equally important for all industrial activities or for all activities. If you do textile, if you do agro processing, it may not be as important, you may not need such a London type city or such a big city to capture the agglomeration or the economies of scale needed for that activity to flourish. Similarly, there is actually quite a few other reasons why we may observe metropoles, which have to do with the politics. There may be an urban bias, there may be a metropolitan bias, maybe in the interest of political elites or elites to sort of concentrate, bring industry to more metropoles, because that gives them an opportunity to do some rent seeking by issuing the permits and sort of let them pass by them so that they have opportunities for rent seeking. And finally, also get congestion of course, as things get more concentrated, there's also congestion, it's very hard to estimate that as well. But by and large that literature would sort of predict, would say, you get big agglomeration economies that would sort of accelerate growth. Now we come from the other side. Secondary towns, you come sort of from the rural non-farm employment literature. Secondary towns, smaller towns are known to be important as mediators for the flow of goods and services between the rural hinterlands and the larger cities. And sort of they basically are important for generating rural non-farm employment. And rural non-farm employment is important, which has been documented to be important for poverty reduction. Now again, why would that then link with sort of what could be reasons why the poor may find it easier to find jobs if these rural non-farm jobs have to be generated. Is that better for them to be generated in small towns or big cities? Sort of one could think of a theory, sort of a Harris-Todaro type of theory where you say, look, in the cities there may well be higher wages, the expected wage in the city is higher. But the likelihood that you're actually going to get a job may be much slower. So if everybody wants to be an actor, we all go to Vegas, and in your favorite actor, you may indeed get, if you become that actor or actress, you hit the jackpot. There are many who never get there. And they get to Vegas, but never become the famous actor. So high average wage, which attracts the migrants, but also higher unemployment and thus sort of slump formation, et cetera. The other argument would be that it's actually much easier for migrants, for rural farmers, to find a job which matches their skills. So they may actually find it easier to get to that place, to the job. And because they can keep connections with the hinterland, they can keep their land tenure issues, they keep access to their land, they maintain their social ties, so migration costs are less. So there are a number of reasons why it may be easy for poor people to actually get to the rural town where, if they arrive there, they may then have a higher likelihood or an equal likelihood of getting a job. So it's sort of the probability of finding a job once you're there and the probability of getting there, which we call, the latter one, we call the size effect, and we'll talk a little bit about it. But these rural towns may well lead to lower agglomeration economies and thus a slower growth pattern, that sort of a flipside of the argument. I'm giving you two arguments which are sort of from within, from within the city, from within the rural town. But of course, these activities, they sort of have externalities, they have effects on the hinterlands. And it may well be that if you have the big city, it's sort of vibrant, that pulls the whole hinterland in and creates demand for the goods and services produced in the hinterlands. So the rural farm employment is really driven by the demand coming from the big city. Now, again, is that effect larger for a city or for a number of spread-out secondary towns which each have that effect on their hinterland? It may not be as strong, but in this case, the coverage of the area may be much larger. These are sort of some ways to think about why a small town versus a big city may have differential effects on the pace of poverty reduction. But ultimately, these are empirical questions. So that's what we'd like to talk or to look into today. We're going to do this in two ways. First, we'll have a scale study and we'll do some cross-country regressions so looking at the cross-country experience. We're going to do very simple. We basically divide the population in three groups. The people in agriculture. So if rural people in agriculture, then you have people who are in non-agriculture. Some of them will be rural and then some of them will be in secondary towns, urban. And then the third group will be the people who live in cities larger than one million. This couple of is simply chosen by the data, is determined by the data. We have basically, in practice, we'll just divide the population in three groups. We take the people who are employed in agriculture from their rural data. We take the people living in the big cities, one million plus cities, from UN World Urbanization Prospects and the rest is the missing mill. These are the people in the mill, which is sort of residual. So as I said, we'll do a first simple case study from Kagera and then look at the cross-country experience. In the cross-country experience, what we'll try to do is we'll basically do sort of a simple growth, poverty to growth elasticity with a little small twist on it. So this is the change in poverty. This is your growth. So if, now we look at where people move to, out of agriculture, they move out of agriculture into that middle or they move out of agriculture into the metropoles and the rate at which that happened would not have an additional effect on poverty reduction. So controlling for growth and some other usual controls, country fixed effects, et cetera, does it matter where people go to for poverty reduction is what we'll test eventually. But let's first tell some stories from Kagera. Kagera is a province in the north of Tanzania. There was a survey done which was started in 91, 94 over that period, so about 20 years ago. And the households were revisited in 2010, so about 15 years later. What's quite unique about this survey is that not only were the households revisited, if they went back to the household and the household wasn't there, they would actually go and trace that household. They would trace all the individuals in that household. So they would go back, even if the household was there, but let's say the daughter and the son had left, they would actually go and visit, see where these people had come through. So it's in a way, it's tracking individuals, so it's a panel of individuals. We have about 3,300 individuals' households, so the amount of attrition was quite small. The co-author Joachim DeWerty actually lives in that area and has been running the survey from there. Now, the area, broadly speaking, is, if you look at the broad socioeconomic indicators, it's not so different from the rest of Tanzania. That said, this by no means meant to be representative as such for Tanzania, but there are also some peculiarities to that region. Now, let's just look a bit at what happened to these sort of dynamic pairs. Where did these people move to? People were in farming or were in the middle, and then they moved from the farm to the farm. So in 1991, in that period, they were a farmer, and they were a farmer as well in 2010. Or they moved either to rural unemployment or to a small town, or they moved to the city, in this case, the Aras Alam, Mwanza, or even in Uganda, Kampala. This is the total sample, basically about 82% in 1991-1994, so at the beginning were farmers. That dropped substantially to 48%. So that's sort of what we have here. We clearly have a big move out of farming. Let's look at what happened to consumption of these people. Are we surprised? Going to the city pace. Is there incomes for those who moved to the city increased by triple, basically, increased by 233%. Those who moved to the middle also saw their income go up quite a bit, less, but it doubled. My God, those who stayed in farming, only 61% increased. It's not so much. I will go to the city. That's what we would think. That's sort of how we think about these things, and here I want to do a little thought experiment for you. I was reading the book, Thinking Fast and Thinking Slow, so you'll decide for yourself if you're a fast or a slow thinker. Daniel was pretty frustrated with himself because he felt that his statistical intuition was not really up to par, and he thought that it was him. So basically he started out testing this a bit to make sure what was happening, or whether this was him, was his problem, or whether it was a more widespread problem. So one of the experiments he talks about in the book, he says, you know, if you have this person, sort of an introvert person, talk to too many people, sort of always on his own, what do you think, what is more likely? Do you think that type of person is more likely to be a librarian, or do you think that type of person is more likely to be a farmer? You decide, we'll come back. Come back to this story here. So basically here we go. Very fast growth rate, but if you look at the contribution to total consumption growth in that sample, actually it was the middle who contributed 42%. The farmers, they both had the sort of the same, while their growth rate was much less, they still contributed 18%, and the farmers moving to the city, they also contributed an 18%. So a big difference here, a fast growth rate, but the real contribution came from people moving to the middle. Let's go to poverty. What happened in terms of poverty? Yes, people who moved to the city, they left poverty. No poverty left. But look at this number here. 434, there were 945, about 30 percentage point, 28 percentage point reduction in poverty, and basically you get almost half of the people who left poverty, who exited poverty, were people who moved to that middle. Also here 300 out of 945. So one in three farmers almost left exited poverty by staying in farming, and only one in 713 over 945. One in seven people exited poverty by moving to the city. Go back to your thought experiment, what was more likely, actually it might well be more likely to be a farmer, because there are many more farmers than there are librarians in the population, and even though there are only very few of these are likely to be that introvert type, as a whole you may still be more likely to be a farmer. Just want to bring out this idea here, what's driving this is the share in the population. Just many more people. Even the farmers, one in three exited poverty, there's just many more of them. Then people go moving to the city. Now so what do we take away from this thought experiment? So almost one in two individuals move out of poverty by moving to the middle. Only one out of seven did so by moving to the city. And in this, there's no economic, it's a simple descriptive table. We don't abstract from how the city may have affected growth in the rural, interlent, and vice versa. The other thing when I say looking at the likelihood, sort of the likelihood of finding a job, there is sort of a sign that these are the number of percent of people who are unemployed, and so that rate is slightly higher than those who move to the middle or were farmers. I've only done half of the presentation, so I need to go fast now and get to the punch line. So basically, I now look at this equation. So controlling for growth, does it matter where people move to? So we're basically looking at whether these coefficients are the same. So basically whether it matters where you go out of agriculture into the middle or out of agriculture into the metropolitan. This is the baseline result. Basically, there is growth, reduces poverty, the $1 a day, $2 a day, and there's an additional effect of moving to that middle on the rate of poverty reduction. We do a whole series of robustness tests to these results which we can talk about later on. They're shown here. We also look whether it's more likely that it is the size effect or the Harris-Tiberou effect. We do find more sign of the size effect that it's easier for people to actually likely more... It's easier for them to move to that place as opposed to once they're there that they have a higher likelihood of finding a job. So the unemployment rates are not that different across these jobs. But I want to sort of... This is still all controlling for growth. So you may well say the agglomeration effect works through the growth channel and you only have a lagged effect in the next round. So one simple way of testing that is by just re-running this regression without growth and you still see this effect of the share of the middle dominating so that it seems to work a lot moving to that middle as an important factor. Two final points. To look at this idea a bit further, is it indeed the case? We look at the Gini. We look at is there a correlation between living in the middle and the share of people in the metropoles and the Gini coefficients. And basically you find countries with a higher metropolitan share of the population have higher Gini. So there's sort of an idea, it works to increase the inequality channel. At the same time, the flip side of this that remember the agglomeration economies we do indeed find this is a growth regression, a simple one, but there is sort of signs of a correlation here as well between being in that change rate of the metropolitization if you wish, sort of has a higher effect on growth than a move of people to the middle. This is my last slide. To conclude here, what we sort of want to bring to the table with this paper is basically shift the conversation a little bit. That urbanization is one thing but we need to go beyond that. We need to think about the nature of the urbanization that may well matter for the pace of poverty reduction. So sort of in this paper, sort of cross-country experience as well as some case studies that migration out of agriculture in that middle that may well be associated with faster poverty reduction than agglomeration in the mega cities. A big question of course if you were to believe this, how do you do it? That's for the next talk. Thank you. Thank you very much.