57712 WO R K IN G PAP E R N O .12 Cities: Engines of Growth and Prosperity for Developing Countries? Gilles Duranton WORKING PAPER NO. 12 Cities: Engines of Growth and Prosperity for Developing Countries? Gilles Duranton © 2008 The International Bank for Reconstruction and Development / The World Bank On behalf of the Commission on Growth and Development 1818 H Street NW Washington, DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org www.growthcommission.org E-mail: info@worldbank.org contactinfo@growthcommission.org All rights reserved 1 2 3 4 5 11 10 09 08 This working paper is a product of the Commission on Growth and Development, which is sponsored by the following organizations: Australian Agency for International Development (AusAID) Dutch Ministry of Foreign Affairs Swedish International Development Cooperation Agency (SIDA) U.K. Department of International Development (DFID) The William and Flora Hewlett Foundation The World Bank Group The findings, interpretations, and conclusions expressed herein do not necessarily reflect the views of the sponsoring organizations or the governments they represent. The sponsoring organizations do not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of the sponsoring organizations concerning the legal status of any territory or the endorsement or acceptance of such boundaries. All queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World Bank, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522- 2422; e-mail: pubrights@worldbank.org. Cover design: Naylor Design About the Series The Commission on Growth and Development led by Nobel Laureate Mike Spence was established in April 2006 as a response to two insights. First, poverty cannot be reduced in isolation from economic growth--an observation that has been overlooked in the thinking and strategies of many practitioners. Second, there is growing awareness that knowledge about economic growth is much less definitive than commonly thought. Consequently, the Commission's mandate is to "take stock of the state of theoretical and empirical knowledge on economic growth with a view to drawing implications for policy for the current and next generation of policy makers." To help explore the state of knowledge, the Commission invited leading academics and policy makers from developing and industrialized countries to explore and discuss economic issues it thought relevant for growth and development, including controversial ideas. Thematic papers assessed knowledge in areas such as monetary and fiscal policies, climate change, and equity and growth and highlighted ongoing debates. Additionally, 25 country case studies were commissioned to explore the dynamics of growth and change in the context of specific countries. Working papers in this series were presented and reviewed at Commission workshops, which were held in 2007­08 in Washington, D.C., New York City, and New Haven, Connecticut. Each paper benefited from comments by workshop participants, including academics, policy makers, development practitioners, representatives of bilateral and multilateral institutions, and Commission members. The working papers, and all thematic papers and case studies written as contributions to the work of the Commission, were made possible by support from the Australian Agency for International Development (AusAID), the Dutch Ministry of Foreign Affairs, the Swedish International Development Cooperation Agency (SIDA), the U.K. Department of International Development (DFID), the William and Flora Hewlett Foundation, and the World Bank Group. The working paper series was produced under the general guidance of Mike Spence and Danny Leipziger, Chair and Vice Chair of the Commission, and the Commission's Secretariat, which is based in the Poverty Reduction and Economic Management Network of the World Bank. Papers in this series represent the independent view of the authors. Cities: Engines of Growth and Prosperity for Developing Countries? iii Acknowledgments Comments and feedback from an anonymous referee, Gustavo Bobonis, Bob Buckley, Vernon Henderson, Frédéric Robert-Nicoud, Cam Vidler, and especially Patricia Annez and Keith Head were very much appreciated. iv Gilles Duranton Abstract This paper reviews the evidence about the effects of urbanization and cities on productivity and economic growth in developing countries using a consistent theoretical framework. Just like in developed economies, there is strong evidence that cities in developing countries bolster productive efficiency. Regarding whether cities promote self-sustained growth, the evidence is suggestive but ultimately inconclusive. These findings imply that the traditional agenda of aiming to raise within-city efficiency should be continued. Furthermore, reducing the obstacles to the reallocation of factors across cities is also desirable. Cities: Engines of Growth and Prosperity for Developing Countries? v Contents About the Series ............................................................................................................. iii Acknowledgments ..........................................................................................................iv Abstract .............................................................................................................................v 1. Introduction ..................................................................................................................1 2. A Simple Graphical Framework to Think about Urban Development ................2 3. What's Special about Cities in Developing Countries? .......................................16 4. Do Cities Matter for Growth?...................................................................................32 5. Policy Conclusions.....................................................................................................43 References .......................................................................................................................45 Cities: Engines of Growth and Prosperity for Developing Countries? vii Cities: Engines of Growth and Prosperity for Developing Countries? Gilles Duranton1 1. Introduction Urban policy interventions in developing countries often have two objectives. The first is to make cities "work better" by improving their provision of local public goods, from sewerage to public transport. The second is to limit urbanization, the movement of people from rural areas to already crowded cities. This dual agenda is driven by the idea that the priority for policy should be to alleviate the grim life of urban dwellers in developing countries and slow down the growth of cities to prevent more misery. Although there is no doubt about the abysmal conditions in the slums of Nairobi or Calcutta, is the gloomy outlook of many governments in developing countries about their cities justified? More precisely, we ask two related questions. First, do cities favor economic efficiency? Second, do cities and urbanization bolster self-sustained growth? To answer these two questions, an integrated and consistent theoretical framework is first developed. We start from the idea that the entire urban system is an equilibrium outcome (arguably one where politics and other institutional features play a fundamental role) and lay down a simple graphical device to describe the main feedbacks. The framework is then expanded to focus on a number of specific features of cities in developing countries. This highly tractable and flexible framework is also used to interpret the existing evidence about cities and urbanization in developing countries. To the first question about whether cities foster (static) economic efficiency, the answer from the literature is a resounding yes. Cities provide large efficiency benefits and there is no evidence that they systematically hurt particular groups. We show below that this result provides support for the first pillar of traditional 1 Gilles Duranton holds the Noranda chair in International Trade and Development in the Department of Economics at the University of Toronto. A Philip-Leverhulme Prize winner in 2003, his research interests are both theoretical and empirical. On the theory side, he is interested in the modeling of urban system and the micro-foundations of agglomeration economies. His empirical work is concerned with the measurement of location and concentration in continuous space, the estimation of urban increasing returns and the identification of spatial externalities. He served as consultant on regional and urban policy for various European governments and international organizations. Cities: Engines of Growth and Prosperity for Developing Countries? 1 urban policies (those that seek to improve the functioning of cities). The importance of efficiency benefits from cities also suggests that restricting urbanization entails losses. Our theoretical framework also underscores key complementarities in urban policy and cautions us about a number of pitfalls. The second question about the dynamic benefits generated by cities is more difficult to answer. The evidence suggests that cities can favor economic growth provided the largest city in a country does not grow too large compared to the others. Although this evidence is not strong enough to provide the basis for radical policy initiatives, it raises further doubts about policies that take a negative stance on cities and discourage labor mobility. The priority for policy should be to prevent or curb the worst imbalances in urbanization rather than attempt to slow it down or reverse it. Broadening the focus from within-city efficiency to between-city efficiency even suggests that reducing the obstacles to the reallocation of factors and activities across cities is a highly desirable policy objective. In conclusion, there is nothing wrong with the first traditional pillar of urban policy in developing countries, although it may not be for the reasons that are commonly alleged. In addition, instead of restricting the influx of people into the cities, the second pillar of urban policies in developing countries should be to favor the mobility of resources across cities and regions, while avoiding their concentration in only one primate city. The rest of this paper is organized as follows. Our graphical framework is presented in section 2. This section also discusses the main policy issues in the framework. Section 3 reviews the empirical evidence about greater economic efficiency in cities. This section also expands the framework to discuss urban features that are salient in developing countries such as primate city favoritism and dual labor markets. Section 4 focuses on the evidence about the effects of cities on the dynamic of growth and development. Finally, section 5 provides further discussion of a number of policy issues and offers some conclusions. 2. A Simple Graphical Framework to Think about Urban Development Modeling Cities Economic theories concerned with cities have a common underlying structure.2 This structure contains three elements: A spatial structure, a production structure, and some assumptions about the mobility of goods and factors. These elements are necessary for any model of cities to be well specified. Spatial structure. Since cities are located somewhere, some description of geography is obviously needed. It is often convenient to distinguish between the 2 The material in this subsection is adapted from Combes, Duranton, and Overman (2005). 2 Gilles Duranton internal geography of cities and their external geography. Internal geography is concerned with land, housing, infrastructure, and internal transport. External geography is about the development of new cities and how cities are located relative to each other and to the location of natural resources.3 Production structure. It may be tempting to specify an aggregate production function that directly relates primary factors to the final output, as is customary in much of economic analysis. This standard simplification is often not adequate in our context because cities are characterized by increasing returns to scale and how such increasing returns are generated has potentially important policy implications. In particular, detailed assumptions are needed about labor, the nature of products, the production function of individual firms, the input-output structure that links firms, and how the latter compete. Three main mechanisms can be used to justify the existence of urban increasing returns (Duranton and Puga, 2004). First, a larger city allows for a more efficient sharing of indivisible facilities (such as local infrastructure), risks, and the gains from variety and specialization. For instance, a larger city makes it easier to recoup the cost of an infrastructure or, for specialized input providers, to pay a fixed cost of entry. Second, a larger city also allows for a better matching between employers and employees, buyers and suppliers, partners in joint- projects, or entrepreneurs and financiers. This can occur through both a higher probability of finding a match and a better quality of matches when they occur. Finally, a larger city can facilitate learning about new technologies, market evolutions, or new forms of organization. More frequent direct interactions between economic agents in a city can thus favor the creation, diffusion, and accumulation of knowledge. This typology of sources of urban increasing returns differs from the traditional Marshallian "trinity" (Marshall, 1890), which talks of spillovers, input-output linkages, and labor pooling. Marshall's typology is about "where" agglomeration effects take place (market for labor, market for intermediates, and a mostly absent market for ideas) whereas the one discussed above is about the type of mechanism at stake (sharing, matching, learning). Hence, these two typologies complement each other since the three mechanisms highlighted above (and their associated market failures) can take place in different markets. Good policies will require knowing about both the type of market failures at play and the markets where they take place. 3 Depending on the focus of the analysis, some aspects need to be explained in great detail while others can be modeled in a very simple fashion. For instance, models that emphasize market access often propose a detailed modeling of the external geography of cities. On the contrary, models that focus on housing supply usually assume a very simple external geography but need to pay more attention to the internal geography of cities and the micro issues related to the operation of land markets. Furthermore, both the internal and external geography of cities are often taken as exogenous. This may be true in the short run, but this need not be the case in the long run as distances within and between cities can be modified following changes in policy or technology. Cities: Engines of Growth and Prosperity for Developing Countries? 3 Hence, the first general feature that emerges from the literature is that many different mechanisms can generate urban increasing returns. The second main feature highlighted by the literature is that sources of urban increasing returns are also sources of urban inefficiencies. For instance, specialist input producers in a model of input-output linkages may not be remunerated for increasing the choice of inputs in a city. In a matching framework, firms are not compensated for increasing the liquidity of their local labor market. With learning spillovers, workers are not rewarded for the knowledge they diffuse around them. More generally, private and social marginal returns do not in general coincide in a city. This means that urban production is inefficient, in the sense that it does not make the best possible use of local resources. These two features have important implications. The pervasiveness of market failures hints at a strong role for policy. However, the appropriate corrective policies depend on the exact mechanism at play. The corrective policies associated with urban knowledge spillovers are not the same as those correcting for imperfect matching on the labor market. Given that many mechanisms generate similar outcomes, identifying the precise sources of urban agglomeration and their associated market failures is extremely difficult (Rosenthal and Strange, 2004). In terms of policies, this suggests extreme caution when trying to "foster agglomeration effects." From a modeling perspective, the fact that a variety of mechanisms can generate urban increasing returns is very good news because we expect agglomeration economies to be a robust feature of cities. This also suggests that we can assume the existence of urban increasing returns without having to rely on a specific mechanism. Mobility of goods and factors. Assumptions about mobility, both within and between cities, play a crucial role. These assumptions need to cover the geographical mobility of goods, services, primary factors, ideas, and technologies. The extent to which material inputs and outputs are tradable clearly varies across sectors. Among primary factors, land is immobile, although its availability for different uses (for example, housing versus production) is endogenous. Capital is often taken as highly mobile, with (roughly) the same supply price everywhere. As emphasized below, the (imperfect) mobility of labor, both geographically and sectorally, is a fundamental issue that warrants careful treatment. Finally, the mobility of ideas and technologies determines how production varies across space. The "3.5-Curve" Framework of Urban Development We now present a simple model of a city in an urban system. This model, in the spirit of Henderson (1974), can be represented diagrammatically. The wage curve. The first key relationship is the city aggregate production function relating total output in a city to its inputs. If the three primitive factors of production are land, labor, and capital and if furthermore land is perfectly immobile while capital is perfectly mobile, the focus of our attention needs to be 4 Gilles Duranton on labor. Rather than considering output per worker as function of the size of the urban labor force, it is technically equivalent, but more fruitful in terms of interpretation, to focus our attention on an inverse-demand for labor that relates the wage of workers to the size of the urban labor force. This curve is represented in figure 1 (a) and referred to as the wage curve in what follows. Figure 1: Baseline Case: A Typical City w N Wage Curve wC (a ) wB N HB HC Cost of Living Curve (b) H N w N H N wB HB B Labor Supply A C Curve wC HC (c ) Net Wage Curve N NB NC Cities: Engines of Growth and Prosperity for Developing Countries? 5 In figure 1 (a), the wage in a city is increasing in the size of the urban labor force, reflecting the existence of urban agglomeration externalities. The intensity of urban increasing returns is measured by the slope of the wage curve. Since the nature and intensity of increasing returns is expected to differ across activities, so will the exact shape of the wage curve. This upward-sloping wage curve stands in sharp contrast with "neoclassical" wage curves that slope downwards. Urban increasing returns have received a considerable amount of theoretical attention. Modeling cities in this way is consistent with a fundamental stylized fact. Most, if not all, measures of productivity per capita increase with city size (see below for a discussion of the evidence in developing countries). In turn, a higher productivity in larger cities can explain why a disproportionate share of economic activity takes place in a small number of places rather than spreading uniformly over space as would be predicted by a neoclassical model. If anything, the level of the wage curve (for any level of employment) is even more important than its slope. The concentration of employment fosters urban productive efficiency. However, human concentration is not the only determinant of urban efficiency, which also relies on a broad range of infrastructure from roads and international airports to well-functioning rental markets for commercial property. Hence, the wage curve can differ across cities because of differences in infrastructure and local institutions. Level differences for the wage curve can also occur because of natural endowments and a set of other factors discussed below. As also made clear below, differences in the wage curve also naturally lead to cities of different sizes in equilibrium. The cost of living curve. The second relationship relates the costs of living in a city to its employment size. The main components of the cost of living are the cost of commuting, housing and other consumption goods. It seems reasonable to assume that commuting costs increase with population because a larger population implies longer commutes and more congested roads. Similarly, one expects increasing population to drive up the cost of land and thus, of housing. Under some conditions to be clarified below, a larger city with a higher cost of land also implies higher retail costs and thus a higher price for consumption goods and other nontradables. In figure 1 (b), the cost of living in a city is increasing in the size of the urban labor force, reflecting increasing urban crowding.4 For reasons that will become 4 An important technical issue needs to be mentioned. An increase in productivity, which raises local wages, may be expected to have a positive effect on the demand for land and thus on its price. If commuting is paid in units of time, higher wages also lead to a higher shadow cost of commuting. Hence an upward shift in the wage curve implies a downward shift in the cost of living curve on the figure. We can ignore these two issues by assuming that the cost of living is paid in monetary terms only and that housing consumption per household is fixed. It is important to note that more formal modeling either ignores these effects or suggests they are second order and thus do not completely offset the direct effect of a shift to the wage curve. Hence, to keep the exposition simple, we ignore these effects in what follows. 6 Gilles Duranton obvious, this curve is drawn with a reversed Y-axis. The precise shape of the cost of living curve is driven by the details of the specific mechanisms that underpin it and is ultimately an empirical matter. However, that the cost of living should increase with population is intuitively obvious. As discussed below, the empirical literature strongly supports this notion. Beyond its shape, the level of the cost of living curve is also of fundamental importance. First, just like with the wage curve the cost of living curve is also riddled with market failures. For instance, unpriced urban congestion implies an inefficiently high cost of living for any level of population. Poorly defined property rights can also prevent the efficient densification of cities since investors may be reluctant to invest in property upgrading when, for example, they face a risk of expropriation. Second, a low cost of living in a city also relies on a vast number of local public goods. In this respect, the provision of roads and public transport to ease commuting is important. The provision of many other public goods of a less capital-intensive nature such as security or air cleanliness also matters. Like the wage curve, the cost of living curve is also expected to differ across cities because the latter differ in their shape, availability of land, and so forth. The net wage curve. If we think of the wage curve as representing labor market earnings and of the cost of living curve as the expenditure associated mainly with housing and commuting, the difference between the wage curve and the cost of living curve is represented in figure 1 (c) by the net wage curve.5 On that figure, this difference is bell-shaped. This corresponds to the case where agglomeration economies dominate crowding costs for a small population, while the reverse occurs for a large population. For this to be the case, the wage curve must be steeper than the cost of living curve before a certain threshold and flatter beyond. At this threshold, net wages reach their peak (point B in the figure). This peak can be interpreted as identifying a "pseudo-optimal" city size, which maximizes net wages per capita in the city.6 The reason this is only a "pseudo- optimum" (also called a constrained optimum) rather than a true optimum is due to the existence of market failures in production and in the cost of living. These market failures imply that, on the figure, the wage and cost of living curves are not as high as they could be. 5 This curve is only a difference between two other curves and thus cannot count as a independent relationship. Hence, the "3.5 curve" name for this framework. Rather than measuring the cost of living curve in units of the numéraire, one could view the cost of living as a price index. In that case, using the difference between the two curves would only be warranted when a using a log scale for both the wage and the cost of living curves. Alternatively, the net wage curve might be modified to represent ratio of the other two curves. 6 With no natural obstacle to city creation, it is reasonable to aim for the maximization of per capita surplus rather than total city surplus. Note also that we implicitly assume that labor is the sole factor of production since we equate the surplus accruing to labor to total surplus. It is easy, albeit cumbersome with our graphical framework, to consider other factors of production. Cities: Engines of Growth and Prosperity for Developing Countries? 7 The labor supply curve. The second curve represented in figure 1 (c) is an inverse labor supply curve. For any level of net wage, it indicates the amount of labor supplied in the city. For simplicity, we assume that labor supply is a function of city population and ignore labor force participation decisions. In that case, this curve essentially captures the migration response to the wage in the city under consideration. A flat labor supply curve, as in the figure, implies perfect mobility. In a fully urbanized country, labor mobility takes place primarily across cities and the labor supply curve of city mainly reflects the conditions in other cities. In a country not yet fully urbanized, labor mobility mostly implies rural-urban migration and the labor supply curve of a city mainly reflects the conditions of rural hinterlands. We return to this important issue below. Note finally that city-specific effects, such as amenities, shift this curve. More attractive cities face a labor supply curve that is below that of less attractive ones. This is because workers accept a lower net wage and are compensated by higher amenities. Equilibrium. The equilibrium of the model in absence of any policy intervention can now be derived. The intersection between the labor supply and net wage curves determines the equilibrium. It corresponds to a situation where workers obtain the net wage they require to come to and stay in the city. This intersection between these two curves may not be unique. In figure 1 (c), the two curves intersect twice (at points A and C). The labor supply curve first cuts the net wage curve from above (at point A) and then from below (at point C). Point A is not a stable equilibrium. It is easy to see that a small positive population shock raises the net wage. In turn, from the supply curve, this higher net wage attracts more workers, which again raises net wages and this process continues until the city reaches point C. By the same token, a negative shock leads population and wages to fall to zero. Turning to the second intersection at point C, a similar argument verifies that this equilibrium is stable. From figure 1 (c), once we have established the equilibrium population in the city, NC, we can trace upwards to figures 1 (a) and 1 (b) to read off the equilibrium wage, wC, and cost of living, HC, respectively. Before turning to welfare and policy issues, three important points must be discussed. First, note that, to the extent that agglomeration effects take place within sectors, cities have a tendency to specialize. To see this, it is useful to consider two hypothetical activities in a city. These two activities are entirely unrelated and each has its own productivity curve and a given initial level of employment. Workers in both activities face the same cost of living since everyone is competing for the same land. On the other hand, the two activities offer, in general, different wages. Then, workers are expected to leave the activity with the lowest net wages and move to the other. This movement happens only 8 Gilles Duranton when the city is specialized in a single activity.7 More generally, it is inefficient to have "disjoint" activities in the same city since they bring no benefit to each other and crowd each other's land market. We thus expect the economic composition of cities to reflect this. Hence, should agglomeration effects take place mostly within sectors, cities should be specialized. If, instead, agglomeration effects take place at a broad level of aggregation with strong linkages across sectors, more diversity should be observed. Second, different wage curves (and cost of living curves) lead naturally to cities of different sizes. A higher wage curve for a city implies a higher net wage curve and, in turn, a larger equilibrium population. Similarly, cities specialized in sectors with stronger agglomeration economies will also reach a higher equilibrium population. See Duranton (2007) and Rossi-Hansberg and Wright(2007) for models in which technological shocks on the wage curve generate realistic distributions of city populations. Finally, it is important to note that the analysis of cities is inherently a "general equilibrium" problem, in which the researcher has to look beyond the direct effect of a change and assess the induced changes that follow. Doing this is possible only if there is a clear analytical framework within which the various effects interact. Welfare in the 3.5-Curve Framework To discuss policy, we proceed in stages. This subsection discusses the main welfare issues. This discussion should be viewed more as way to reach a deeper understanding of our framework than a practical policy guide. General policy issues are addressed in the next subsection before turning to specific policy problems in a development context in section 3. Uncompensated externalities in production. The first source of inefficiencies stems from the production structure itself. As argued above, the microeconomic foundations of the increasing returns operating inside cities are all associated with market failures. First, the indivisibilities at the heart of sharing mechanisms generate a number of inefficiencies. Like all indivisibilities, they imply that only a limited number of players enter the market. This results in imperfect competition and the (socially inefficient) exploitation of market power. If new entrants increase the diversity of, say, local inputs, they are unlikely to reap the full benefits of this increase in diversity. We also expect firms to make their entry decision on the basis of the profits they can make rather than the social surplus they create. Under imperfect competition, this is again inefficient. Second, with matching mechanisms, a different set of market failures is at play. For instance, firms neglect the positive effects of their vacancies on the job search of workers. Finally, there are also many possible market failures associated with learning 7 Should, for some unspecified reason, the two activities have the exact same returns, a small employment shock, positive or negative, in any of the two activities again creates a small asymmetry between the two activities and leads again to full specialization. Cities: Engines of Growth and Prosperity for Developing Countries? 9 mechanisms. Under imperfect intellectual property rights protection, firms are likely to invest too little in knowledge generation. In absence of rewards for knowledge diffusion, too little of it takes place. Firms in cities may also be reluctant to train their workers if, for example, they expect them to be poached by competition in the future. These are only several of the inefficiencies that can occur when production takes place under increasing returns. If these inefficiencies were suppressed, wages would increase in the city for any level of employment. Starting from the thin wage curve in part (a) of figure 2 (i), solving for the inefficiencies in production leads to the thick line in part (a) of figure 2 (ii). Uncompensated externalities in cost of living. The second source of market failures is related to the cost of living curve. If the private marginal costs paid by residents were equal to social marginal costs (that is, the costs to the economy), there would be no inefficiency in cost of living. With no congestion, a perfectly functioning land market, and redistribution of the land surplus, this equality between private and social marginal costs holds naturally. Empirically, we expect none of these three assumptions to be satisfied: land markets are subject to significant frictions and are strongly regulated through planning and zoning regulations; increases in land values are not taxed away; and, as cities get more crowded, congestion becomes more important. About the latter, note that traffic congestion is a major form of congestion in cities, but by no means the only one. Most local public goods, from parks to cultural events, and many amenities are also subject to negative congestion externalities. Poorly defined property rights over urban land also constitutes a critical issue in many developing countries. The main implication of congestion and frictions on the land market is that the cost curve in the absence of corrective policy is distorted. With proper corrective policies, it should be possible to reduce costs of living for any population level in the city. For instance, a congestion tax would reduce the level of traffic congestion in the city and can increase total surplus. Starting from the thin cost of living curve in part (b) of figure 2 (i), fixing the inefficiencies in cost of living leads to the thick line in part (b) of figure 2 (ii). A higher wage and a lower cost of living imply a higher net wage curve in part (c) of figure 3 (ii). After curing the market failures in production and cost of living, the net wage curve and the labor supply curve intersect at points D and F (rather than A and C prior to the policy interventions). The net wage curve has its maximum at point E instead of B. Just like A, point D indicates an unstable equilibrium. The only stable equilibrium is in F. This new equilibrium offers a higher net wage than the one with no intervention at point C. Population is also higher. This is because solving for the inefficiencies in production and cost of living makes the city more attractive. In turn, the labor supply response implies that workers migrate to the city. 10 Gilles Duranton Figure 2: Welfare Effects w N w N Wage Curve Wage Curve (a ) (a ) N N Cost of Living Curve Cost of Living (b) (b) Curve H N H N w N H N w N H N E F B B C C A A D Labor Supply Curve Labor Supply (c ) Curve (c ) Net Wage Net Wage Curve Curve N N (i) Initial situation (ii) Fixing market failures in production and cost of living w N w N Wage Curve Wage Curve (a ) (a ) N N Cost of Living Cost of Living Curve Curve (b) (b) H N H N w N H N E w N H N E F G G (c ) (c ) Labor Supply Labor Supply Curve Net Wage Curve Net Wage Curve Curve N N (iii) Making labor perfectly mobile (iv) Solving for the city coordination failure Cities: Engines of Growth and Prosperity for Developing Countries? 11 The extent to which a higher net wage curve leads to a higher population versus a higher net wage depends on the slope of the labor supply curve. Perfect mobility (that is, a flat supply curve) implies that all the gains from curbing the inefficiencies in production and cost of living are translated into a higher population and more crowding. In absence of mobility, a vertical labor supply curve implies that the upward shift of the net wage curve leads only to higher net wages. It is also important to note that after solving all the inefficiencies associated with production and cost of living, the equilibrium in F does not coincide with the first best at point E. Barriers to migration. The third source of inefficiencies is related to the labor supply curve and thus to the migration process. The labor supply curve is driven by two different sets of forces. First, it echoes the net wage in the rest of the economy. For many developing countries, we expect the labor supply curve to be mostly a reflection of rural earnings. In that case, a higher net wage in rural areas implies a higher labor supply curve. Second, barriers to migration are also reflected in the labor supply curve. More costly mobility implies a higher and steeper labor supply curve. Eliminating obstacles to mobility in part (c) of figure 2 (iii) thus leads to a lower and flatter labor supply curve. As a result, the equilibrium shifts to point G. Interestingly, this new equilibrium implies a larger population and a lower net wage than the previous situation at point F. The net wage decreases because reducing barriers to mobility makes it easier for newcomers to settle in the city. Since the city at point F is already in the region where the marginal agglomeration gains are dominated by the marginal losses in urban crowding, an influx of newcomers lowers the welfare of existing residents. This negative result underscores a fundamental policy issue. Urban economies are second-best economies. Nothing guarantees that fixing a market failure always brings a city closer to optimality. Solving for the market failures in the wage and cost of living curves as well as removing barriers to migrations is not enough to lead a city to its first-best efficiency. There is yet another market failure that prevents cities from reaching their optimal size. Unless this last market failure is also fixed, reducing the barriers to mobility may not improve welfare in a city.8 The city coordination failure. As made clear by part (c) of figure 2 (i), the equilibrium with no policy intervention (point C) is not efficient and is located to the right of the pseudo-optimum (point B). Without any corrective policy, existing cities are too large with respect to their pseudo-optimum so that employment concentrates into too few cities that are too big. The reason behind this inefficiency is a coordination failure. Fixing the inefficiencies embedded in the wage, cost of living, and labor supply curves changes nothing to the city coordination failure. In part (c) of figure 3 (iii), the 8Even though welfare in the city under consideration decreases, aggregate welfare increases. This point is made clear below. 12 Gilles Duranton equilibrium size, point G, is still inefficiently large compared to the first-best in E. It is easy to understand why this inefficient situation can be sustained. No one wants to move alone and develop a new city because it would mean forming a very small and thus very unproductive city. It is worthwhile to move to a new city only if it is already large enough or if a big enough group of workers and firms decides to coordinate their move. The creation of such a new city would be desirable for everyone since existing cities would become smaller and thus be able to offer higher net returns. The problem is of course that, in absence of corrective policy (or market for cities), there is no mechanism to coordinate the movement of workers to new cities. To solve this governance problem and to reach the first-best in E, two solutions can be envisioned. First, the city under consideration may directly restrict its population size to reach point E. Doing so implies rejecting residents and sending them to places where they are worse off. Depending on where these rejected residents go, this can increase the cost of living in other cities or increase rural population and thus arguably lower agricultural earnings. Hence, this first solution is a partial equilibrium response to the city coordination failure that generates negative general equilibrium effects. The second alternative is to create new cities and coordinate the move of residents to these new cities. This creation of new cities implies a reduction of population for previously oversized incumbent cities and thus an improvement in welfare for their remaining residents. Should new cities be populated by rural migrants, this would also imply a decrease in rural population and thus arguably an increase in agricultural earnings. In turn, a higher welfare outside the city implies a higher labor supply curve. In this case, the general equilibrium effects are positive.9 New cities can then be created until the labor supply curve hits the net wage curve at point E. At this stage the entire urban system is fully efficient. Practical Policy Considerations It is now time to take a more practical look at urban policies. A fundamental policy question should first be answered: Is it worth it for policy to bother about cities at all? As just argued, cities are riddled with market failures: production is inefficient, congestion is rife, and overcrowding is expected to be the rule. The above welfare analysis also makes clear that full urban efficiency is extremely demanding to achieve. Hence, there is a strong temptation to view the "urban problem" in developing countries as an unmanageable pathology and neglect 9 More generally, general equilibrium effects (that is, what happens outside the city) matter and play a fundamental role. Changes taking place outside the city under consideration affect the labor supply curve and thus its equilibrium. These interdependencies can mean that a worsening of the situation outside the city (that is, a lower labor supply curve) leads to an influx of new residents and a worsening of the welfare in the city as well. The importance of general equilibrium effects also implies that improving the functioning of only one city makes it grow but has ambiguous implications regarding welfare in this city. A better-functioning city becomes attractive and the new residents can crowd out all the gains. Cities: Engines of Growth and Prosperity for Developing Countries? 13 cities. That would be wrong. Having numerous inefficiencies only implies that cities are much less efficient than they could be and that there are important gains from well-designed urban policies. Furthermore, existing urban inefficiencies do not imply that cities are less efficient than their rural alternatives. Actually, the very success of cities in developing countries points to the opposite. However suboptimal they may be, cities typically offer higher returns and better long-term opportunities. Neglecting cities and restricting their access can only have negative consequences: a worsening of urban inefficiencies and "overcrowded" rural areas, which in turn implies low returns to agriculture and an exacerbation of rural poverty. Going into the details, two important points need to be made about the wage curve. First, the wage curve reflects a considerable number of evolutions that are determined well beyond the city under consideration. To take a simple example, many developing countries have policies that distort agricultural prices relative to manufacturing prices. Since cities in developing countries are specialized into manufacturing and services, any increase in relative manufacturing prices is likely to translate into higher urban wages and thus a higher wage curve. In turn, this should lead to larger cities. More generally, national technology and government policies are going to be reflected in the wage curve of any particular city, affecting its level and, sometimes, its slope.10 The second issue with the wage curve is related to the market failures beneath it. To repeat, the existence and growth of cities is driven by a variety of mechanisms whose relative importance is extremely difficult to identify empirically. The market failures associated with these mechanisms then all require different corrective policies. For instance, corrective policies aimed at dealing with labor market matching problems have nothing to do with those aimed at, for example, fostering knowledge diffusion. Put differently, we need some corrective policies for inefficiencies that we know close to nothing about. This suggests some caution.11 Given the limited possibilities for policy to raise the wage curve, the cost of living curve is a more promising area of action for city governments. The main reason is that many of the key determinants of the cost of living curve such as traffic congestion are reasonably well-identified problems that we know quite a bit about. From sewerage to public transport, there are many components of the cost of living curve for which local governments can make a big difference. The second main policy issue related to the cost of living curve has to do with poorly 10 Regarding the latter, think for instance about modern telecommunication technologies that may affect the intensity of agglomeration effects. 11 Furthermore, it is also the case that these market failures are likely to occur in all cities. Creating a more efficient labor market or favoring the diffusion of knowledge is more appropriate for central rather than local governments. The main tool for local governments with respect to the wage curve should then be the provision of productive local public goods. However, a complete discussion of this issue, including of the qualifications that apply to the preceding statement, would take us well beyond the scope of this paper (for recent reviews, see Helsley, 2004; Epple and Nechyba, 2004). 14 Gilles Duranton defined property rights and the inefficient operations of the land market. This salient issue in many developing countries is dealt with at greater length below. Finally, note that many other policies of local governments such as the provision of public goods and amenities also get reflected into the cost of living curve. This only reinforces the point that the cost of living curve is the traditional area of expertise of city governments and should remain so. Turning to labor mobility, it is clear that a flatter labor supply curve can potentially lead to important welfare gains by allowing workers to move from low net wage areas to high net wage cities.12 As hinted above, this increase in mobility is best carried out by central governments since any city that unilaterally increases labor mobility may decrease its welfare. This prescription of greater labor mobility runs contrary to many policies in developing countries that aim instead at restricting internal migrations. Given the importance of labor mobility to improve efficiency in the short run, and possibly to foster economic growth in the long run as well, this issue is further developed in the next two sections. The last prescription of the framework regards the fact that cities tend to be too large in equilibrium, which calls for the creation of new cities and the coordination of their settlement. This is of course a practical minefield and this recommendation should be taken with extreme caution. Past experiences of city creation, and more particularly of capital city creation, in developing countries have often led to mixed results (or worse). While in the United States new cities are often created by private developers (Henderson and Mitra, 1996), few developing countries appear to be able (or willing) to follow suit. Besides, developing countries already appear to host many very small cities. Hence, rather than creating new cities, the challenge is the lack of growth of many small cities.13 These issues are discussed at greater length below. 12 An important technical caveat applies here. In absence of pure externality in the wage and cost of living curves, it is always good from an efficiency (and welfarist) perspective to have workers move from low-wage (rural) areas to high-wage cities. This result holds even though cities are in a region of decreasing returns. This occurs because the difference between the net wage curve and the labor supply curve exactly measures the social marginal gain of one more migrant in the city. This is no longer true in the presence of pure externalities. Then, a new worker into the city can raise the wage of all other workers (through agglomeration effects) but also increase their cost of living. If the increase in cost of living associated with the externality is very large, the private gains from the move for the migrant and the higher wages for all workers can be more than offset by the cost of living loss of all the other inhabitants. Given how big spatial disparities can be in developing countries (Aten and Heston, 2005), the congestion externalities would need to be extremely large for migration from poor to rich areas not to raise overall output. This case remains to be made empirically. 13 One may object to this and argue that all cities, small and big, are already expected to be oversized. There is no contradiction here if one acknowledges that cities should grow with their pseudo-optimal size. Note further that the growth and industrialization of small cities is all the more important since developing countries often have their international comparative advantage in mature manufacturing sectors. Small and mid-size cities are natural locations for such activities (Henderson, 1997). Cities: Engines of Growth and Prosperity for Developing Countries? 15 3. What's Special about Cities in Developing Countries? Empirical Support for the Framework Before going deeper into policy issues, let us first discuss our framework in light of the empirical evidence in developing countries. In brief, the literature offers support for all its main building blocks: an upward-sloping wage curve, costs of living rising with city size, a bell-shaped net wage curve, and some labor mobility driven by net wage differentials. Starting with the wage curve, there is a large literature that documents the existence of agglomeration economies in developed economies (see Rosenthal and Strange, 2004, for a review). The main conclusion of this literature regards the existence of scale economies of 3 to 8 percent (that is, a 10 percent increase in the size of an activity in a city raises productivity in this activity by 0.3 to 0.8 percent). These agglomeration effects take place both within sectors (localization economies) and between (urbanization economies). Although there is far less research about agglomeration economies in developing countries, the results are usually similar. As in developed countries, studies of agglomeration economies in developing countries regress some productivity outcome in cities (and sectors) on city measures of economic activity within or across sectors. See Rosenthal and Strange (2004) and Combes et al. (2008b) for more details about this type of methodology. Following Henderson (1988)'s study of localization economies in Brazil, several studies have found quantitative evidence of localization effects. For instance, Henderson, Lee, and Lee (2001) find localization economies for Korean industries, more particularly traditional industries. Lall, Shalizi, and Deichmann (2004b) for India, and Deichmann et al. (2005) for Indonesia provide similar evidence, albeit less strong. Further evidence about localization effects can be found in a number of case studies looking at a wide variety of countries and sectors (see Overman and Venables, 2005, for references). There is also evidence of urbanization economies in developing countries. Henderson, Lee, and Lee (2001) show that they matter for advanced sectors in Korea. There is also evidence of urbanization economies for India. It is rather weak in Lall, Funderburg, and Yepes (2004b) but much stronger in Lall, Koo, and Chakravorty (2003). Deichmann et al. (2005) also find mild evidence of urbanization effects in Indonesia for a number of sectors. The results of Au and Henderson (2006a,b) about Chinese cities are also consistent with a mix of localization and urbanization economies. This literature is further discussed in Henderson (2005) and Overman and Venables (2005), who provide detailed reviews of agglomeration findings for developing countries. Strong localization economies are expected to foster the growth specialized cities while strong urbanization economies foster that of diversified cities. 16 Gilles Duranton Evidence of both localization and urbanization economies is consistent with the existence of diversified cities and specialized cities in developing countries.14 Two main criticisms can be made to these studies. First, they usually do not control for the individual characteristics of workers (observed and unobserved). It could be that measured agglomeration effects only reflect the sorting of more productive workers in bigger and more specialized cities rather than true agglomeration economies. Using French data, Combes, Duranton, and Gobillon (2008a) show that such sorting is empirically important and goes a long way towards accounting for observed spatial disparities. Nonetheless, controlling for sorting does not make agglomeration effects vanish. Second, most of the available findings concern primarily the formal sector. Household surveys that cover both the formal and informal sectors will hopefully be more widely used in the future. At this stage, we can only note that the linkages between formal and informal sector firms are often intense, which suggests that agglomeration effects are generated within both the formal and informal sectors with benefits that accrue to both. It is also worth mentioning that the case-study evidence that supports the existence of agglomeration effects also strongly supports the idea that the informal sector is a strong contributor. Turning to the cost of living curve, the evidence is scarce. Early works by Thomas (1980), Henderson (1988), and Richardson (1987) show a fast rise in the cost of living with city size. These cost of living findings are confirmed by more recent work from Henderson (2002a) who looks at a broader cross-section of cities. He finds the elasticities of various cost of living measures to cities size to be between 0.2 and 0.3.15 Finally, Timmins (2006) develops a novel methodology to infer the "true cost of living" from widely available data using a model of location choice. He implements his approach on Brazilian data and finds that the cost of living increases with city size above a certain threshold.16 The evidence about the net wage curve is thin. The main difficulty here is that, with sufficient labor mobility, we expect all cities to be on the decreasing portion of the net wage curve following the stability argument exposed above. Having most cities on the decreasing portion of the net wage curve is, for instance, consistent with the Brazilian findings of da Mata et al. (2007). Direct evidence about net returns to size being bell-shaped is provided by Au and 14 Despite strong evidence about localization economies, it seems that there are few specialized cities in many developing countries relative to the United States. Other factors such as high transport costs must thus be invoked to explain these weak patterns of urban specialization. See below for more on that. 15 These elasticities, just like those for the wage curve, are estimated for observed (that is, equilibrium) sizes. It is then unsurprising that the cost-of-living elasticities with respect to population are higher than the wage elasticities. This is exactly what the framework predicts should happen in equilibrium. 16 Interestingly, he also finds that the cost of living also decreases with population below the threshold. This suggests that the cost of living is high in large cities and in small and isolated places. Cities: Engines of Growth and Prosperity for Developing Countries? 17 Henderson (2006a,b) for cities in China. They exploit the fact that the Chinese government has imposed strong barriers to labor mobility that have constrained urban growth.17 As a result, a steep labor supply curve is expected in China. Provided it is steep enough, some cities can be too small in equilibrium and a bell-shaped net wage curve can be estimated. Interestingly, Au and Henderson (2006a,b) find that Chinese cities tend to be significantly undersized. This results in large income losses. The other finding of Au and Henderson (2006b) is that the net wage curve is quasi-flat after its maximum. This suggests that cities may become grossly oversized under free mobility but that the costs of being oversized are small (unlike the costs of being under-populated). The migration mechanism that underlies the labor supply curve has been widely studied. Greenwood (1997) proposes a general survey of internal migrations in developed and developing countries while Lall, Selod, and Shalizi's (2006) review focuses on developing countries. A key finding of the literature is that internal migration flows in developing countries are consistent with an upward-sloping labor supply curve. Representative of this literature, Brueckner (1990) and Ravallion and Wodon (1999) find that the direction of migration flows is consistent with existing differences in net wages. In their work in Bangladesh, Ravallion and Wodon (1999) also address the slope of the net wage curve by showing that there are persistent differences in living standards across areas despite the absence of formal barriers to mobility. Closer to the spirit of the labor supply curve in our framework, da Mata et al. (2007) estimate a population supply function for Brazilian cities and find the elasticity of population to income per capita to be between 2 and 3. This is quite elastic, but still far from perfect mobility. Barrios, Bertinelli, and Strobl (2006) show that, in sub-Saharan Africa, there is a direct link between climate, which directly affects living standards in rural areas, and urban growth. This type of finding is consistent with an important role for shocks that shift the labor supply curve, up or down (Poelhekke, 2007). It also suggests that in less advanced countries, the labor supply curve is mostly driven by living conditions in the countryside rather than in other cities.18 This, in itself, is consistent with the traditional notion of surplus labor (Lall, Selod, and Shalizi, 2006). The final conclusion that can be drawn from Barrios, Bertinelli, and Strobl (2006) is more subtle. There is a negative correlation between urban growth and the welfare of urban dwellers. This negative correlation may explain why many developing country governments attempt to restrain urbanization. However, this correlation 17 Although the absolute numbers for Chinese cities look impressive, urbanization has proceeded at a slow pace compared to many other countries during their industrial takeoff. For the recent past, the growth of Chinese urban population has been 3-4 percent per year as opposed to 5-6 percent for Korea, Indonesia or Brazil at a comparable stage. In cross-country analysis, China's urbanization rate is very low relative to its level of GDP (Henderson, Quigley, and Lim, 2007). 18 The corollary of this is that worsening rural conditions, which lower the labor supply curve, lead to "urbanization without growth," as documented for instance in Fay and Opal (1999). 18 Gilles Duranton is not causal. Negative agricultural shocks lower the labor supply curve and workers flock to the cities thereby lowering urban welfare. Hence, cities can still offer efficiency benefits despite a negative correlation between urban growth and urban net wages. On the contrary, preventing rural dwellers from moving to the cities will make them worse off. We now turn to the development of new cities. Although the theoretical literature has recently made progress on efficient city development (Henderson and Venables, 2006), little is known empirically about it. Using data about world cities spanning several decades, Henderson and Wang (2007) use a 100,000 population threshold to track the entry of "new cities." Several interesting findings emerge. First, in a typical country, the rate of growth in the number of cities is not statistically different from that of population growth. This suggests that new cities do indeed rise and the rough proportionality between the entry of new cities and population growth is reassuring. Nonetheless, this does not say much about the efficiency of the process of city creation beyond ruling out the notion that it is entirely dysfunctional. Henderson and Wang (2007) also show that the emergence of new cities is favored by democratization and government decentralization whereas it is slowed down by having a large fraction of educated workers. With world urban population growing by about a hundred million per year, there is no doubt that those issues deserve further attention. Finally, there is scant evidence about cities being too large in general. Very strong barriers to labor mobility have made Chinese cities too small according to Au and Henderson (2006a,b). There is unfortunately no other study that attempts to look at this question without making heroic assumptions about what optimal city size is. In light of the framework exposed above and its predictions about cities being oversized, casual observation of cities in developing countries raises some apparently puzzling facts with respect to city size. While many megacities in developing countries such as Karachi in Pakistan with a population well above 10 million are arguably "too big," most cities in developing countries are much smaller. In Thailand for instance, there is only one "large" city with a population above 300,000. How is it that Bangkok could be too big with a population nearing 6 million while the fifth largest city in Thailand, Chiang Mai, could also be too large with a population only around 150,000? To solve this puzzle, we extend our framework into two directions to consider the issue of primate city favoritism and that of market access. Primate City Favoritism Urban primacy is a well-known feature of urbanization in developing countries (see, for example, Henderson, 2005). To explain why in so many developing countries the largest city is often disproportionately larger than the second largest, the literature has focused on two arguments. First, urban primacy is sometimes attributed to protectionist trade policies. In the model of Krugman and Livas Elizondo (1996), trade liberalization reduces urban primacy because it Cities: Engines of Growth and Prosperity for Developing Countries? 19 allows all cities to import differentiated goods from abroad. In turn, this equalization of market potential reduces the tendencies for the agglomeration of manufacturing in a single core city. Although correct, their model is arguably very particular. Rather than equalizing market potential, it seems more reasonable to assume that trade liberalization gives privileged market access to coastal cities or to cities close to trading partners. Then, inland primate cities can obviously see their dominance reduced by trade liberalization. Mexico City since NAFTA, which served as motivating example to Krugman and Livas Elizondo (1996), may be an illustration of this. On the contrary, coastal primate cities can see their dominance reinforced by trade liberalization (Fujita and Mori, 1996). Think of Buenos Aires in Argentina whose primacy remained unabated despite trade liberalization. The effects of trade policy are thus theoretically ambiguous. Consistent with this theoretical ambiguity, the empirical support for trade- based explanations of urban primacy is weak. Studies that find a negative effect of trade on primacy often do so because they fail to control properly for other channels that can influence primacy and are correlated with trade (see, for example, Moomaw and Shatter, 1996). The better and more recent studies (Ades and Glaeser, 1995; Nitsch, 2006) suggest that trade plays no systematic role with respect to urban primacy. Instead, political and institutional factors appear to be at the root of the primacy phenomenon. There is strong evidence of a positive association between unstable and undemocratic regimes and urban primacy (Ades and Glaeser, 1995; Davis and Henderson, 2003). The exact underlying mechanism(s) is nevertheless not fully elucidated. The story is often told in terms of dictatorial regimes bribing the residents of the primate city because they are afraid of being overthrown by social unrest. Direct evidence about this mechanism is lacking. Furthermore, this type of explanation implicitly assumes fairly strong state institutions able to tax their countryside and redistribute the proceeds to the primate city. On the contrary, it may be argued that undemocratic and unstable regimes are weak and favor primate cities "by default." Primate city favoritism can work through a myriad of small decisions from underpriced gasoline and better provision of local public goods to better business opportunities for government cronies in the primate city (Henderson and Becker, 2000; Henderson, 2002a,b). In this respect, the many regulations and permits that govern economic activity in most developing countries could play an important role. Being close to a center of power makes it easier to obtain permits or to circumvent their necessity. A complementary explanation points at a better road infrastructure linking the primate city to the rest of the country (Saiz, 2006). We can thus speak of primate city favoritism, but keeping in mind that this favoritism takes place in many different ways. Primate city favoritism can readily be incorporated into our framework. We assume for simplicity that favoritism (or the lack thereof) affects primarily wages. Earnings are higher than they would otherwise be in the favored cities. 20 Gilles Duranton They are also lower than they would otherwise be in the other cities since favoritism comes at a cost for them. Part (a) of figure 3 represents the higher wage curve for the favored city and the lower wage curve for a nonfavored city. As a result, with the cost of living curve being the same in both cities on part (b) of the same figure, the net wage curve of the favored city is above that of the nonfavored city. It is then easy to see that the equilibrium size of the favored city is larger than that of the nonfavored city in part (c) of figure 3.19 Because of general equilibrium effects, the labor supply curve is also lower than it would be in absence of favoritism in part (c) of figure 3. The potentially large misallocation of resources associated with primate cities suggests that some effective policies to reduce urban primacy are needed.20 However, dealing effectively with this problem is going to be hard. First, primate city favoritism manifests itself in many different ways and there is no definite evidence at this stage about which channel(s) matters most. Interestingly, the Korean experience hints that administrative deregulation may be a powerful tool to reduce urban primacy (Henderson, Lee, and Lee, 2001). Red tape may be costly for all businesses but more so for those located far away from the main center so that deregulation is more beneficial to them. Second, the political economy associated with urban primacy may be very difficult to break. Cronies who benefit handsomely from their proximity to the political power are unlikely to easily accept a leveling of the playing field. Third, the theoretical findings of Henderson and Venables (2006) suggest that governments may play a role in anchoring expectations about which secondary cities will get developed. Their development may then alleviate primacy. However, anchoring expectations about future urban development may be subject to time-inconsistencies and subject to an inefficient political economy. Internal Market Access The proposition that a good access to markets matters can be traced back at least to Harris (1954). It was recently revived by Krugman (1991) and the ensuing work. This body of work is referred to as the New Economic Geography and is summarized in Fujita, Krugman, and Venables (1999), Baldwin et al. (2004), and Combes, Mayer, and Thisse (2008c). 19 On the graph, the favored city is larger but not disproportionately larger than the nonfavored city. This for clarity only. With a flatter downward sloping portion of the net wage curve, this difference can be made much bigger. According to Au and Henderson (2006b), the net wage curve is empirically rather flat beyond it maximum in China. 20 One may argue that such policies have been attempted for a long time. This is true but many of these policies, such as the relocation government activities, did not provide the right incentives for residents to relocate and took place in a framework of highly controlled labor mobility. Cities: Engines of Growth and Prosperity for Developing Countries? 21 Figure 3: Primate Cities Favoritism w N Wage Curve wB (favored city) Wage Curve (nonfavored city) (a ) wA N HA (b) HB Cost of Living Curve H N w N H N wA HA A B (c ) wB HB Labor Supply Curve Net Wage Curve N Our minor adaptation of Krugman's (1991) model considers two regions and two sectors. Agriculture produces a homogenous good under constant returns in the hinterland of each region. For simplicity, this good is assumed to be perfectly tradable and is produced by immobile workers. In each region, there is a city where manufacturing firms operate under increasing returns. Each monopolistically competitive firm employs mobile manufacturing workers to produce a separate variety of differentiated product, which is demanded by 22 Gilles Duranton consumers in both regions. Manufacturing varieties are costly to transport between regions so that firms' sales have a "home-market bias." The wage of manufacturing workers is determined as follow. Consider a "high" level of transport costs. This is a reasonable assumption for most developing countries.21 Due to high transport costs, manufacturing producers in each city are partly insulated from imports from the other region. Producers in a city can thus charge high prices, which in turn imply high manufacturing wages locally. If manufacturing expands, the regional market gets more crowded. This happens because, although the expansion of manufacturing also implies a larger urban market, the size of the local market (that is, the whole region) does not increase proportionately (remember the fixed agricultural sector in the hinterland). Furthermore, with high transport costs, very little of the increase in manufacturing output gets exported. Basically, when transport costs are high, manufacturing wages decrease with the size of the urban manufacturing workforce. This alone would lead to a downward sloping wage curve and a complete dispersion of manufacturing. However it seems difficult to completely write off the efficiency effects of urban agglomeration described above. This suggests that the wage curve is determined by opposing forces: market access versus agglomeration economies. Put differently, as a city grows larger, its regional market gets more crowded so that the prices of locally produced goods decline (and this, in turn, leads to lower urban wages) but it also becomes more efficient (which raises urban wages). Assume that market access effects dominate agglomeration effects in small markets while the reverse holds in large markets. This implies a wage curve that first slopes downwards and then upwards. This wage curve is represented in figure 4 (a). We concentrate on this case because it has more interesting implications than its opposite. To defend it empirically, one could argue that negative market crowding effects can be very strong at the margin in a very small market while they are going to be much milder if many firms are already operating in a market. Furthermore, it could also be argued that a minimum city size is needed for agglomerations economies to take place. It is important to understand that transport costs affect not only the wage curve (through the production of goods) but also the cost of living curve (through their consumption). A small isolated city may face very low housing and commuting costs. However, consumption goods can be very expensive because most of them are produced elsewhere and need to be shipped to this city at a very high cost. As the city grows and produces more manufacturing, the price of goods declines since a smaller proportion of them need to be imported. 21Further details about this case and a complete explanation of the low transport cost case can be found in Combes et al. (2005). Interestingly, the tradeoff between the two main forces described below is resolved differently when transport costs are low. Cities: Engines of Growth and Prosperity for Developing Countries? 23 Figure 4. Internal Market Access w N Wage Curve wC (a ) wA N HA (b) HC Cost of Living Curve H N w N H N B Net Wage Curve (c ) Labor Supply wA HA A C Curve wC HC N NA NC On the other hand, other components of the cost of living such as housing and commuting increase with city size. Again, we have forces pushing in opposite directions. It seems reasonable to assume that higher housing and commuting costs eventually dominate when cities grow very large. This suggests that the cost of living first decreases and then increases as cities grow. The cost of living curve corresponding to this situation is represented in figure 4 (b). 24 Gilles Duranton Subtracting the cost of living from the wage implies that the net wage first decreases, then increases, before decreasing again with city size.22 In figure 4 (c), the net wage curve and the labor supply curve intersect three times. Ignoring the unstable equilibrium in the middle, we are left with two stable equilibria. Cities are either very small, at point A, or much bigger, at point C. The optimal city size (point B) is somewhere in between. Compared to figures 1 and 3, the novelty in figure 4 is the existence of small cities in A whose growth is limited by strong crowding on the product market and insufficient agglomeration effects. This crowding is in turn caused by high transport costs and the difficulty for the cities to export their output. The representation in figure 4 is important because it provides a strong rationale for the coexistence in many developing countries of small stagnant cities and large primate cities. High costs of trade between cities may also explain why cities in developing countries may not be as fully specialized as in developed countries. Urban specialization makes little sense when the costs of intercity trade are very high. The literature offers strong empirical support regarding the importance of market access for cities in developing countries. Using two different approaches, Lall, Koo, and Chakravorty (2003) and Lall, Funderburg, and Yepes (2004b) underscore the importance of market access in India. Strong effects of market access are also found in Brazil (Lall, Funderburg, and Yepes, 2004a; da Mata et al., 2007) and Indonesia (Deichmann et al., 2005; Amiti and Cameron, 2007). This within-country evidence is complemented by a large literature that looks at the importance of market access at the country level (Redding and Venables, 2004; Head and Mayer, 2004). Regarding the shape of the cost of living curve, the evidence is much thinner. In large part, this is due to the general paucity of research on this issue. However, the paper that currently defines the frontier on the topic (Timmins, 2006) finds strong evidence for Brazilian cities of nonlinear cost of living curves taking the shape hypothesized above. Let us now turn to policy implications. Improving market access implies better access to other markets but it is also synonymous with a loss of protection for local firms. Depending on which effect dominates, the wage curve can shift upwards or downwards. Better market access for small isolated cities also implies a less steeply decreasing wage curve so that a flatter wage curve (at least in its early part) is expected. With better access, we also expect a lower cost of living. On balance, for small cities better market access implies a flatter and possibly higher net wage curve. In turn, this implies that the small city equilibrium at point A should shift to the right (city growth) or even disappear 22This requires the wage curve to decrease faster initially than the cost of living as the city grows. We expect this to happen because manufacturing wages are expected to decline proportionately to the price of local manufacturing goods whereas the cost of living in the city is expected to decrease less than proportionately since the prices of the agricultural good and imported manufacturing are unchanged while the other components of costs of living increase. Cities: Engines of Growth and Prosperity for Developing Countries? 25 entirely keeping point C as the only stable equilibrium. With broad-based gains from better market access, we can also expect a higher labor supply curve through general equilibrium effects. As a result of a higher labor supply curve, the equilibrium size of large cities would decrease. The final outcome could be smaller big cities but a larger number of them. In practice, market access is improved by two sets of policies. The first is about building and developing roads and other transport infrastructure such as airports or high-speed train lines. The second is about removing impediments to trade across regions, from administrative hurdles to cartelized distribution networks. A number of caveats must be kept in mind. First, much remains to be done because most existing specifications in the empirical literature are not derived directly from theory (Head and Mayer, 2004, 2006). Put differently, the importance of market access is established but it is still unclear how it precisely works. Next, the development of road networks may have perverse effects. Linking small cities to large economic centers increases the market potential of the former but it may increase that of the latter even more and thus reinforce primacy instead of reducing it. The U.S. experience nonetheless suggests that there can be large productivity gains associated with the development of an integrated transport network (Fernald, 1999). Finally, improving market access may also have some effects at a geographical scale greater than cities. A key prediction of modern regional economics is that lower levels of transport costs can lead first to increased regional agglomeration, and then possibly decreased regional agglomeration for even lower levels of transport costs (Fujita, Krugman, and Venables, 1999; Combes, Mayer, and Thisse, 2008c). However, with better transport infrastructure, there is a possibility of a group of winning cities in core regions and a group of cities left behind in the periphery.23 One could think of coastal Chinese cities versus hinterland cities or high-plateau cities in Colombia versus cities on the Colombian Caribbean coast, and so forth. In summary, urban primacy is often attributed to a dysfunctional political economy leading to primate city favoritism. There is a lot of empirical support for this. A complementary explanation points at high internal trade costs leading to either large or small cities. A lot of the evidence is consistent with this explanation as well. In both cases a reduction in urban primacy is desirable. Doing it through a reduction in primate city favoritism may be effective but is hard to implement politically. Improving market access for isolated cities may be politically easier to achieve but the precise effects of better access are more difficult to predict since improved access may reinforce primacy. 23The theoretical ambiguities about the effects of transport costs on regional agglomeration are analyzed in depth by Baldwin et al. (2004). See also Fujita and Mori (2005) for a systematic study of how transport costs can affect cities in a regional setting. 26 Gilles Duranton Migration and Dual Labor Markets The framework developed above assigns a positive (and equalizing) role to internal migrations and labor mobility. This is in contrast with some of the academic literature and much of the policy reality in developing countries. From internal passports in China and the "nativist" policies of Indian States to the resettlements policies carried on in Africa and Latin America, there is a strong bias against free labor mobility in many developing countries. As shown above, restricting the movement of labor is not the right answer if cities become too big as in the framework exposed above. Another justification for antimobility policies rests on the existence of dual labor markets. The argument was first developed by Harris and Todaro (1970) and has been extremely influential in policy circles. Theoretically, it works as follows. There is a formal sector with a fixed number of urban jobs that pay a high wage, wA, in figure 5 (a). In rural areas, workers get lower earnings represented by the labor supply curve in figure 5 (c). This initial earnings gap between the rural and the formal urban sector causes workers to move to the city. Should there be only so many migrants to the city as there are jobs in the formal sector, the city would end up at the social optimum, point A in figure 5 (c). However, when the city is at point A it cannot be in equilibrium because the net wage is above that received in rural areas. If there are more workers than available jobs in the formal sector, the model assumes that jobs are randomly attributed among the city residents. The lucky ones get a job in the formal sector while the unlucky ones get a job in the informal sector of the same city. This informal sector offers a low wage, w. In this case workers keep moving to the city until the expected wage they receive minus the cost of living (that is, their expected net wage) intersects with the labor supply curve.24 This occurs at point B. It is easy to see that this equilibrium entails cities that are too large. The main difference with the baseline case explored above is of course not that cities are too large--that was the case in the baseline as well--but that it makes sense to curtail entry into the city. Although appealing, the Harris-Todaro argument can be criticized on a variety of grounds (see Lall, Selod and Shalizi, 2006, for an in-depth analysis and empirical references). First, in the model workers end up in the informal sector because of wage rigidities in the formal sector. Trying to solve a problem that occurs in the labor market by restricting the mobility of workers is not the most direct solution and is likely to have a number of unwanted side effects.25 The stark assumptions of the Harris-Todaro model also bias it towards generating overmigration to the cities. For instance, workers in the model are risk-neutral 24 The graph follows the approach of Brueckner and Zenou (1999) who explicitly consider a land market. This already reduces the tendency of cities to become too big compared to the most basic versions of the Harris Todaro model. 25 This rigidity is also partly attributable to a very large and spatially concentrated public sector. Again, restricting urbanization is not the way to deal with a dysfunctional public sector. Cities: Engines of Growth and Prosperity for Developing Countries? 27 and formal sector job are randomly allocated. However, real-life workers are arguably risk-averse and know that formal sector jobs are not randomly allocated so that only those with high chances of getting one are expected to move. The fact that the formal and informal sectors appear quite segmented in most developing countries only reinforces the point. Furthermore, the downward sloping wage and net wage curves predicted by figure 5 do not receive any empirical support as made clear above. All this suggests that the main argument used to restrict labor mobility is relatively weak. Figure 5: Harris-Todaro Migrations w N Wage Curve (formal sector) wA Expected wage (a ) wB w Wage Curve (informal sector) N HA HB (b) Cost of Living Curve H N w N H N wA HA A Net Wage (c ) Curve Labor Supply B Curve wB HB N 28 Gilles Duranton To go beyond a mere rebuttal of Harris and Todaro (1970), we need ask ourselves why restrictions on labor mobility are so widespread in developing countries. A first possibility is to point at an overzealous application of Harris and Todaro (1970) by policy makers. In such a case, policies can change after showing the weakness of their underpinnings. Instead, restrictions on labor mobility may be part of a political-economy equilibrium. In this latter case, better policies would then be much more difficult to implement. We need to know more about this issue to understand the nature of the challenge for labor mobility and how it may be overcome. Dual Housing Markets The last key feature of cities in developing countries is the existence of a dual housing sector with a division between the formal sector and squatter settlements (also referred to as slums, invasions, shanty-towns, and so forth). In some large developing country cities, more than half the population live in squatter settlements and face very poor public services provision (if at all), insalubrious living conditions, and a number of constraints associated with the precariousness of their housing. Squatter settlements are often associated with the idea of low-cost and low- quality housing. If it was only this, they would be simply a reflection of the general poverty of some urban dwellers opting out of the formal housing sector because they cannot afford it. Policy decisions regarding what to do with squatter settlements would mostly be choices about how much redistribution to do (or not to do) and whether it is best to do this redistribution through subsidized housing and public services or by other means. These issues are important but there is more to squatter settlements than this. First, it has been widely argued that poorly defined or poorly enforced property rights over urban land, which make squatter settlements possible, could also affect a wide range of other economic outcomes. De Soto (2000) argues that a lack of effective, formal property titles prevents residents of squatter settlements from using their housing as collateral and is thus a major barrier to enterprise development. Although the evidence about the existence of these financial constraints is disputed, Di Tella, Galliani, and Schargrodsky (2007) find that a lack of titles has important effects on the beliefs of people and thus their economic behavior and Field (2007) finds that it also matters for female labor supply. Durand-Lasserve and Selod (2007) provide a detailed review about a broad range of effects associated with a lack of effective titles. Next, squatter settlements may be the outcome of policy distortions. Henderson (2007) argues that binding minimum lot-size is responsible for the growth of squatter settlements in Brazilian cities. Malpezzi (1999) also mentions the prevalence of rent controls that limit the expansion of the rental market, among other things. Cities: Engines of Growth and Prosperity for Developing Countries? 29 Finally, once the absence of public services or their very poor quality is taken into account, squatter settlement may not be so cheap. For instance, water in slums often needs to be bought at a very high price from local water distributors. Without titles, squatters must also often pay a steep price for some form of protection, for example. Taking the last two ideas seriously about exclusionary zoning by the formal sector and the relatively high costs of squatter settlements, it is possible to expand our theoretical framework to gain some insights about dual housing markets. In figure 6 (a), we assume a standard upward-sloping wage curve that applies to all city residents.26 In figure 6 (b), there are three cost of living curves. The dotted curve represents the cost of living in the formal sector in absence of exclusionary zoning. Exclusionary zoning (for example, minimum lot size in Brazil) raises the cost of living in the formal sector, yielding the solid cost of living curve of the graph. The alternative to the formal sector is a squatter settlement. The cost of living in a squatter settlement is represented by the dashed line on figure 6 (b). Because of the high cost of the substitutes for missing public services and other expenses, the cost of living in squatter settlements is higher than in the formal sector in absence of exclusionary zoning. The cost of living in squatter settlements is also higher than the cost of living in the formal housing sector with exclusionary zoning for small cities but lower for large cities. The main justification for this is that the higher cost of "public" services in squatter settlements is rather insensitive to city size whereas the economy in land rent made by squatting is more likely to increase with city size. In absence of exclusionary zoning, the cost of living is always lower in the formal sector and no one would choose to live in a squatter settlement. The net wage curve corresponding to this situation is the dotted thin curve in figure 6 (c). The city equilibrium is reached at point C. With exclusionary zoning, it is cheaper to live in the formal sector than in squatter settlements when the city is small but it gets more expensive when the city grows. The thick line in figure 6 (b) then represents the minimum cost of living under exclusionary zoning. This thick line is solid (that is, the formal sector) for small cities and dashed for larger cities since they expand though squatter settlements. 26In many cities with squatter settlements, a significant proportion of slum dwellers have a job in the formal (production) sector. As a first-order approximation, we assume that all workers benefit from agglomeration effects. A more refined version of the graph would take into account the possibility that slums may often be located far from the main places of work and poorly served by public transport. In such a case, slum dwellers would be less likely to benefit from agglomeration effects. We acknowledge this but this is not core to the argument here which focuses on the housing market. 30 Gilles Duranton Figure 6: Dual Housing Sector w N wB Wage Curve (a ) N Cost of Living Curve (formal sector, Cost of Living Curve no zoning) (squatter settlement) (b) HB H N Cost of Living Curve Net Wage Curve (formal sector, (formal sector, no zoning) with zoning) w N H N Net Wage Curve (with zoning) A B C (c ) Labor Supply Curve Net Wage Curve (squatter settlement) N "Legal" residents Squatters Under exclusionary zoning, the net wage curve is the maximum of the net wage offered either by the formal housing sector or by squatter settlements. It is represented by the thick "continuous and dashed" curve in figure 6 (c). The equilibrium for the city is at point B. Below the X-axis, we can read the city population that resides in the formal sector and in squatter settlements. This analysis suggests a number of policy implications. Imposing regulatory constraints in the formal housing sector may reduce the equilibrium size of the Cities: Engines of Growth and Prosperity for Developing Countries? 31 city but a good fraction of this reduction is crowded out by the growth of squatter settlements. The equilibrium for the city is at point B, and not at point A, as originally intended. Removing unnecessary constraints in the formal housing sector is socially desirable since it lowers the cost of living, hence raises the net wage curve and eliminates squatter settlements. Furthermore, improving the situation in one city only is not enough to raise net wages since a higher net wage curve may keep hitting the same labor supply curve, only at a larger city population. Again, general equilibrium effects matter. "Titling" policies are also desirable because, as argued above, poorly defined property rights have a range of other negative side-effects. After solving disputes over land ownership and the financing of the titles handed-out, the main issues with titling policies are, first, how to avoid further preemptive invasions driven by the expectation of a future title and, second, how to make sure that there is a local tax counterpart to legalized titles. In some respect, the problems in dealing with illegal settlements are the same as those of urban favoritism. There are many dimensions associated with this phenomenon and it is not clear yet which are those that matter most empirically. As with urban favoritism, there is also a political economy of illegal settlements with vested interests that benefit from slums, either directly by charging their residents, or indirectly by providing expensive substitutes for missing public services. These vested interests often pose formidable challenges. 4. Do Cities Matter for Growth? Extending the Framework We now ask whether cities and urbanization can affect the long-run rate of economic growth. To answer this question, we first expand the simple graphical device introduced above to help us think about urban dynamics. In a multiperiod setting, workers are initially endowed with some human capital. Each period, they spend part of their time working. They consume their labor market earnings at the end of each period. They also spend part of each period learning so that at the following period they start with a higher level of human capital and are more productive. Workers work and learn in the city where they are located. We can represent this in figure 7 (a) using two curves. The first is the wage curve used previously. It represents the labor market earnings of a worker depending on the size of the city. The second curve is the wage expansion curve. It represents the (discounted) value of the increase in human capital, again as a function of city size. For brevity, we refer to this curve as the expansion curve. For reasons that will become clear below, it is more convenient to represent this wage expansion as an absolute amount rather than a relative amount.27 At each 27The two key equations of all growth models are the production function and the accumulation equation. The wage curve captures the former while the expansion curve is a version of the latter. 32 Gilles Duranton new period, the wage curve for a given worker in a city shifts upwards as a function of the increase in human capital at the previous period. Before going any further, a number of remarks are in order. First, how much time is spent working and learning is set exogenously. Allowing workers to make decisions about their time allocation would only reinforce the results derived below since workers are expected to spend more time learning where the returns to learning are the highest. Figure 7. Learning in Cities w N Wage Curve wC (a ) VC Expansion Curve N Cost of Living Curve (b) HC H N w N H N B Net Wage Curve A C wC HC (c ) Labor Supply Curve N Cities: Engines of Growth and Prosperity for Developing Countries? 33 Second, since workers entirely consume their labor market earnings at each period, there are no savings. Human capital is the only factor that is accumulated. Third, the expansion curve, which captures the discounted value of learning during the period, is known to the workers. We abstract here from two key complications. The value of what is learnt during the current period may depend on how much is learnt in subsequent periods (as there might be some intertemporal substitutability or complementarity in learning). This value of current learning also depends on future location choices. The extent to which human capital is transferable across cities is discussed below. Turning to the shape of the expansion curve, figure 7 (a) draws an upward- sloping curve. It is less steep than the wage curve. Let us think of this particular shape as a theoretical possibility for now. One could think of three different sets of factors that influence the shape of the expansion curve. First, there are national factors that affect economic growth in the entire country. For instance, institutions or aggregate research and development are often argued to be the main engines of aggregate growth in the literature. When growth is entirely orthogonal to cities, the discounted value of the increase in human capital is the same everywhere. Hence, when location does not matter (as in most of the growth literature), the expansion curve is flat. Then, with a flat expansion curve, the issues of growth and urbanization can be treated separately. Urban issues can be dealt with the tools developed in previous sections whereas the dynamic framework just described boils down to a standard (spaceless) model of economic growth. To understand this result, note that a flat expansion curve at the current period implies that the wage curve shifts upwards at the next period. Following this, the net wage curve also shifts upwards by the same amount. Since the same wage increase takes place everywhere, the labor supply curve also shifts upwards in the same way. As a result, the equilibrium city size is unchanged.28 In addition, with a flat expansion curve, the long-run behavior of the economy depends on how human capital is accumulated. With decreasing returns to the accumulation of human capital, this model is equivalent to a standard Solow (1956) model using human capital instead of physical capital as a factor of accumulation. In this case it is well-known that the economy converges to a constant level of output. With constant returns in the accumulation of human capital, this model then becomes similar to the endogenous growth 28Remember that the cost of living is paid in numéraire and is thus not affected by human capital accumulation. Should the cost of living be paid in units of time (assuming now that commuting requires time and not energy), this benchmark for which cities play no role in growth would look different. More productive labor in cities would raise the opportunity cost of commuting and thus scale the cost of living up. As a result, a complete separation between economic growth and the urban structure is obtained only when the expansion curve is a scaled version of the wage curve. In this case, the equilibrium at the following period would involve scaling the wage and cost of living curves by the same factor. In turn, the net wage and labor supply curves would also be scaled in the same way, leading equilibrium city size to remain unchanged. 34 Gilles Duranton framework of Lucas (1988). The main result is then that a constant positive growth rate of output can be sustained. In this steady state, cities can affect the level of output but not its growth rate. In short, growth takes place in cities but cities do not constitute the engine of growth. Aside from national factors there are also city-specific factors that might be thought to affect the expansion curve in each city individually. As discussed above, the static urban increasing returns associated with an upward-sloping wage curve can be justified theoretically by a number of mechanisms: sharing, matching, and learning. The same mechanisms may be argued to play a similar role with respect to the expansion curve, suggesting that this curve may be upward sloping rather than flat. Learning mechanisms seem particularly relevant here. Actually, the more frequent interactions taking place in cities may not be so much about improving the efficiency of production today but about learning, accumulating human capital, and becoming more efficient in the future. Since Jacobs (1969) and more recently Lucas (1988), this assumption of dynamic increasing returns in cities sustained by some form of human capital externalities or knowledge spillovers has been at the heart of the theoretical literature that views cities as engines of growth (see, for example, Eaton and Eckstein, 1997; Black and Henderson, 1999; Glaeser, 1999; Bertinelli and Black, 2004; Rossi-Hansberg and Wright, 2007). Duranton and Puga (2004) provide a detailed review of how cities can favor the creation, accumulation, and diffusion of knowledge. Let us now explore the implications of the upward-sloping expansion curve of figure 7. In part (b) of that figure, the cost of living curve is the same as in previous figures. Turning, to part (c) the net wage curve now has three components. The "net wage" that workers consider in their location choice for the period is the sum of the wage and the wage expansion minus the cost of living. Since wage expansion is defined as the net present value of the increase in human capital at the current period, these three items are measured consistently. The resulting bell-shaped net wage curve intersects with the labor supply curve at points A and C. Just like in previous figures, only C is a stable equilibrium. The equilibrium wage (wC), the equilibrium value of the increase in human capital (VC), and the equilibrium cost of living (HC) can be read up in panels (a) and (b) of the figure. Part (c) of figure 7 looks similar to part (c) of the baseline figure 1. There is however a fundamental difference because the net wage curve now captures urban efficiency from both a static and a dynamic perspective. The (static) net wage curve associated with the wage curve and the cost of living curve (but excluding the expansion curve) is actually represented by the thin doted curve of figure 3 (c). It is interesting to note that the maximum of the net wage curve for the dynamic model at point B is larger than for the static model. The reason is that a larger city now brings about a higher efficiency at the current period as well as the promise of a higher efficiency in the future. The benefits from Cities: Engines of Growth and Prosperity for Developing Countries? 35 agglomeration are thus stronger than they would be in absence of dynamic effects and optimal city size is thus larger.29 This result suggests that taking a purely static perspective to assess whether cities are oversized is misleading. Lucas (2004) proposes a model of urban-rural migration and learning where rural workers optimally migrate to cities. Early on, their urban wage is very low. In a Harris and Todaro (1970) framework, this would be interpreted negatively as urban unemployment. However, these migrants spend their early time in the city accumulating human capital. This allows them to become more productive later on. Hence with learning, restricting migration to cities has negative dynamic consequences. Turning to the long-run dynamic of our model, note first that the wage curve at the next period sums the current wage curve and the wage increase associated with rising human capital. In turn, a higher wage curve implies a higher net wage curve. If the labor supply curve does not move because of surplus rural labor receiving a constant wage, the equilibrium size of the city increases. Eventually, the exhaustion of surplus rural labor should imply an upward shift for the labor supply curve following the increase in human capital of all workers in all cities. It is also important to note that, because of more learning in larger cities, the wage curve is expected to become steeper as time goes by. Then, a steeper wage curve implies that the maximum of the net wage curve shifts to the right as workers in the city accumulate human capital. In this case, and following Black and Henderson (1999), it is possible to envision the following long-run dynamics under full urbanization. At each period, the net wage grows by the same amount in all cities. This implies that both the labor supply and net wage curves keep shifting up at the same rate. The increase in the slope of the wage curve also implies a relative right shift of the peak of the net wage curve. This evolution is represented graphically in figure 8. This type of dynamics also suggests that the labor supply curve will eventually intersect with the net wage curve at its peak. In this case the learning that takes place in cities leads to both economic growth and the (optimal) population growth of cities, which in turn fuels economic through the expansion curve. So far, to understand the expansion curve, both national and city specific factors have been mentioned. Arguably, interactions between cities (aside from those taking place through the shifts of the labor supply curve) also matter. Before examining them, let us look at the empirical relevance of what has been discussed so far. 29 An extension would be to consider congestion effects specific to learning. For instance, the physical crowding of a city is time-consuming and thus implies that there is less available time to learn and accumulate human capital. In this case, (net) dynamic efficiency no longer monotonically increases with city size. 36 Gilles Duranton Figure 8: Growth in Cities w N Wage Curve (t) Wage Curve (t+1) (a ) Expansion Curve (t, t+1, ...) N Cost of Living Curve (t, t+1, ...) (b) H N w N H N Net Wage Curve (t+1) Net Wage Curve (t) (c ) Labor Supply Curve (t+1) Labor Supply Curve (t) N The Empirics of Growth and Cities Our dynamic framework relies on the existence of some human capital externalities in cities that underpin the upward-sloping expansion curve. These externalities affect the learning of workers. In turn, these assumptions imply that the urban structure affects aggregate economic growth. Let us examine the empirical evidence about these three elements. First, there is a large literature that investigates human capital externalities in cities. Most of the studies look at cities in the United States or other developed countries (Conley, Flyer, and Tsiang, 2003, who look at Malaysia, are an Cities: Engines of Growth and Prosperity for Developing Countries? 37 exception). Here are the key findings (see Moretti, 2004; Duranton, 2006, for surveys). There is a strong association between the average level of human capital in cities and individual wages after controlling for individual characteristics. The effects are relatively large. Estimates of "social returns to education" being of the same magnitude as its private returns are not uncommon. This relationship between city human capital and individual wages is particularly strong when city human capital is measured by the share of university graduates in the city. While these facts are firmly established, the direction of causality is less clear. Nonetheless, there is reasonable evidence that causality runs from city-level human capital to individual wages. Whether existing research has really identified human capital externalities (as opposed to other complementarities) is far less clear because these externalities are notoriously hard to identify empirically. Direct evidence about the channels of transmission of those effects is still missing. In addition, there is also a literature that looks more specifically at learning in cities. Its empirical findings are very suggestive, though again mostly limited to the United States. Glaeser and Maré (2001) show that there is an urban wage premium, which workers retain when they move back to smaller cities or rural areas. Among a number of papers, Peri (2002) and Wheeler (2006) document that wage growth is stronger in cities, particularly for young educated workers. This is consistent with the learning in cities hypothesis. This could also be due to the self-selection of workers with fast career progressions in cities for reasons unrelated to learning. This does not seem to be the case. Freedman (2007) shows that this type of result holds even when controlling for the fact that some workers may experience higher wage growth independently of their location. To conclude, both the literature on human capital externalities and that on learning in cities provide suggestive micro-based evidence about an upward-sloping expansion curve in cities as assumed above. This evidence is nonetheless not entirely decisive and comes nearly exclusively from developed countries. Now, we turn to the relation between the urban structure and economic growth. The empirical growth literature has not been particularly successful at disentangling the causes of long-run growth (see Durlauf, Johnson, and Temple, 2005, for a recent critical review). Sadly, this literature has also barely paid any attention to cities and urbanization as possible determinants of growth. There is indeed only one study, ?, that uses a reasonable cross-section of countries and sound econometric methods to look at the aggregate dynamic effects of cities and urbanization.30 The two main conclusions of that study are the following. First, urbanization per se does not affect economic growth. Second, urban primacy has large effects on economic growth. The first conclusion is rather unsurprising and confirms a broad consensus. Urbanization is a benign transition that, to a large extent, follows the process of development Using the same data, Bertinelli and Strobl (2003) replicate and confirm some of the findings of 30 Henderson (2003) using nonparametric techniques. 38 Gilles Duranton but does not profoundly affect it. The second conclusion is more provocative. Henderson (2003) finds that an increase in urban primacy by one standard deviation (or 15 percent) from the mean (of 31 percent of the urban population living in the largest city) reduces the rate of GDP growth by about 1.5 percent per year. These are large effects. They are also interesting from a policy perspective since urban primacy can evolve relatively quickly. The main issue in any such investigation regards the direction of causality. A strong negative statistical association between urban primacy and growth may not be surprising. A strong causal effect is more so. To deal with causal issues, one ideally needs to find some good instruments for urban primacy, that is, variables that determine urban primacy but are not otherwise correlated with economic growth. One can then use the variation in urban primacy caused by these exogenous variables to assess the causal effect of primacy on growth. Unfortunately, it is hard to think of any variable that would determine primacy and be otherwise uncorrelated with economic growth. In particular, the key determinants of urban primacy, political variables, are expected to have a strong independent effect on economic growth. Instead, Henderson (2003) proceeds as follows. He takes the first difference of all his variables to get rid of any permanent country effects that would be correlated with both economic growth and urban primacy. Then, using a GMM estimation technique, he instruments changes in urban primacy by lagged primacy levels from 10 or 15 years before. This estimation technique yields large effects of urban primacy on economic growth. Economic growth first increases and then decreases with urban primacy. The idea that, for low levels of primacy, a larger main city should foster economic growth is easy to interpret in our framework. Through the workings of the expansion curve, a larger city implies stronger learning for more workers. Provided that somehow there is also a diffusion of this learning to other cities (more on that below), a positive relationship between aggregate growth and the relative size of the main city occurs naturally. The second part of the relationship when the effect of primacy becomes detrimental to growth is more puzzling in light of the above framework. To understand how the framework of figure 7 should be amended to account for this result (just like we amended the baseline static framework above to account for key features of cities in developing countries), we note that Henderson's (2003) results are compatible with two possible interpretations. The first is that the urban structure has a direct effect on economic growth. In this case, we need to consider a situation like in figure 4 where differences in market access lead to two possible equilibria for cities. In this situation with cities that are either small or very big, we now consider an expansion curve that first slopes upwards and then downwards. A downward-sloping section for the expansion curve could be justified by some congestion in learning as the city grows very large. In this case, small cities could be too small for efficient learning to take place while the primate city is too large. With this first interpretation of the Cities: Engines of Growth and Prosperity for Developing Countries? 39 relationship between primacy and growth, promoting growth simply entails reducing the size of the primate city and increasing the size of smaller cities. All cities would then end up on a higher point on the expansion curve. A second possible interpretation of Henderson's (2003) results is that the factors that drive urban favoritism also affect the expansion curve.31 Imagine for instance, that, as suggested above, red tape is at the root of urban primacy. Dealing with government regulations is much easier for firms when they are located in the capital city because of greater proximity with the bureaucrats. Under such circumstances, red tape leads to a higher wage curve for the favored city, which becomes primate as a result. On the other hand, red tape may have a detrimental effect on growth everywhere (albeit less so in the primate city). Hence, red tape may also imply a very low expansion curve leading to very slow growth. With this second interpretation of the findings about primacy and growth, the policy implications differ. A forced reduction of the size of the primate city is unlikely to have much of a dynamic effect since it does not deal with the root cause behind the low expansion curve. Here, the growth problem is not due to primacy itself but to urban favoritism, which causes both primacy and low growth. Although there is scant evidence enabling us to distinguish between the two interpretations, the second one is consistent with the discussion above regarding the strong presumptions of urban favoritism. Diffusion, Mobility, and Growth A key limitation of the approach taken so far is to view each city as an "island of growth." More precisely, it typically assumes that each city can generate economic growth by itself and for itself. Although very little is known about this, especially in a development context, it is fundamental to understand how knowledge flows across places.32 A first line of argument is to recognize that the expansion curve may not be driven by overall city size as in figure 7 but instead by the concentration of "innovation inputs." Duranton and Puga (2001) argue that modern systems of cities experience a division between cities where innovation takes place ("nursery cities" with a very diverse production structure) and cities that are more specialized into the production of one particular set of goods. In developed economies, the last 50 years have also seen a growing separation between 31 To understand why these two explanations are compatible with Henderson (2003), a technical aside is necessary. First, it can be argued that first-differencing by Henderson conditions out all static explanations whereby institutions (a possible missing variable) would explain both long-run growth and primacy. Then, Henderson instruments change in urban primacy by lagged primacy. Past levels of primacy are good predictors of current changes. The key issue is whether these instruments are otherwise uncorrelated with changes in the rate of growth. This is an open question. Although Henderson (2003) shows that overidentification tests are easily passed, time- varying factors closely related to primacy could be at the root of both primacy and growth. 32 Flows of knowledge between developed and developing countries are also fundamental but beyond the scope of this paper. See Keller (2004). 40 Gilles Duranton business centers, which host headquarters and business services, and production cities, which host production plants (Duranton and Puga, 2005). In this type of framework, preventing urban dispersion by favoring a primate city or by slowing down the development of secondary cities prevents the efficient concentration of innovation inputs (scientific personnel, research facilities, and so forth) in nursery cities without too much crowding out from other activities. In turn, this leads to a lower expansion curve in those cities and slower growth there. This also affects secondary cities since the lack of innovation in nursery cities obviously implies a lack of new ideas, new products, and new production processes to be transmitted to secondary centers. This type of claim is of course hard to evaluate empirically. There is good evidence from the Republic of Korea (Henderson, 2002b, 2005) that mature manufacturing quickly moved out of Seoul and relocated to secondary cities after a reduction in urban favoritism. Brazil appears to be following a similar path, albeit more slowly (da Mata et al., 2005). In many other countries, this process of urban change appears to be even slower, if it takes place at all. There may also be a temptation to forcibly concentrate innovation inputs in some cities to create some centers of excellence (or innovation clusters). These policies are unlikely to be successful because of the difficulty of replicating the subtle alchemy of nursery cities. The second line of argument focuses on knowledge flows within countries. The framework developed above relies on human capital being both embedded in the workers (as in the traditional definitions) as well as "in the air" (following Marshall, 1890) in the form of human capital externalities. This is consistent with the notion that knowledge is embedded in people and is acquired by direct contact with "those that know." The corollary of this idea is that flows of people are also flows of knowledge. Hence, more learning, and thus a higher expansion curve, can be achieved through the mobility of skilled labor across cities.33 This argument has been modeled in the context of the mobility of employees between firms (Combes and Duranton, 2006; Franco and Filson, 2006) but not yet between cities. Empirically, Møen (2005) and Freedman (2007) show that technological progress is indeed associated with the movement of skilled workers between firms. Job-hopping appears to be beneficiary to the job-hoppers and to their industry, if not to their employers. In addition, Almeida and Kogut (1999) show that long-distance flows of knowledge, as tracked though patent citations in the U.S. semiconductor industry, coincide with the movement of star scientists across firms in different cities. Interestingly, Agrawal, Cockburn, and McHale 33 Models of urban growth generally ignore this issue. Eaton and Eckstein (1997) is an exception. They assume that human capital accumulation in cities is driven by the "knowledge base" of the city. In turn, the knowledge base of a city is taken to be total human capital in the city plus the discounted sum of human capital of other cities. However, they do not provide a well-specified mechanism regarding these knowledge interactions across cities and only model them as a pure externality. Cities: Engines of Growth and Prosperity for Developing Countries? 41 (2006) also show that the scientists who leave a city keep being cited there. They are gone but not forgotten. To the extent that these findings about highly skilled U.S. workers also apply to highly skilled workers in developing countries, we can draw a number of tentative policy conclusions. First, the general working of the labor market and more specifically the covenants that restrict labor mobility can play an important role to hinder the diffusion of knowledge within and between cities. A lack of labor mobility, especially in the most skilled segments of the labor market, between the main city and secondary cities may be an important contributor to both urban primacy and the backwardness of secondary cities. With limited labor mobility between cities, nearly all skilled labor may go to the main city and stay there. The main city then becomes an island of more advanced knowledge with a much higher wage curve. As highly skilled workers remain in the primate city, their knowledge does not percolate to other cities. These other cities then stay behind technologically and remain small and unattractive because of their low wage curve. A key issue is that, even in absence of formal impediments to labor mobility, this situation can remain since the technological backwardness of small cities may provide little incentive for skilled workers to relocate there. The two-way mobility of skilled labor between cities seems important to foster the diffusion of technologies across places but it may not be the only channel. Although the existing evidence mostly concerns countries and not regions within countries, there is a good case to be made that more trade in goods is also associated with higher growth and convergence across places. In a cross-country setting, Wacziarg and Welch (2003) show that increased openness has large positive effects on growth and investment. Alcalá and Ciccone (2004) also show that the positive growth effects of trade work through total factor productivity. The effects found by Alcalá and Ciccone (2004) and much of the prior literature in a cross-country setting are large. For instance, moving from the twentieth percentile of openness to the median raises productivity by 160 percent according to Alcalá and Ciccone (2004). With no evidence of weaker effects when openness is already high, this suggests that there are potentially large dynamic gains from removing impediments to trade within developing countries. Finally, there is very strong evidence that productivity growth is linked to the process of creation and destruction at the firm level (Davis, Haltiwanger, and Schuh, 1996; Foster, Haltiwanger, and Krizan, 2001; Bartelsman, Haltiwanger, and Scarpetta, 2004). In particular resources need to flow from less to more productive firms and allow new entrants to rise and challenge incumbents. An analysis of this process of reallocation would, of course, go much beyond the scope of this paper. Nonetheless, it is important to remember that in developed countries there is a strong spatial dimension to this process of reallocation as industries tend to change location when their technology evolves (Duranton, 2007). An important conclusion here is that hindering the movement of factors 42 Gilles Duranton across firms, including across firms located in different cities, may have large dynamic costs. 5. Policy Conclusions It is now time to summarize our policy conclusions and consider a number of practical issues regarding their implementation. The first issue is whether a growth agenda leads to urban policy recommendations that differ from those of a traditional agenda, usually more concerned with (static) urban efficiency. The urban efficiency agenda is explored in the first part of the paper. Its main recommendations are the following: eliminate primate city favoritism; improve urban efficiency so as to lower the cost of living curve by dealing with urban crowding, and public good provision; solve the biases that lead to squatter settlements with a reasonable titling policy and urban deregulation; improve market access between cities by developing transport infrastructure and lowering impediments to trade; and do not discourage internal migrations that foster an efficient allocation of the population and have an equalizing effect across places. By underscoring the need for better public service delivery and the importance of housing and commuting issues, this set of recommendations is consistent with some of the objectives of many existing urban policies. The main difference is that our baseline framework also emphasizes labor mobility. This emphasis is strongly at odds with existing urban policies that often seek, on the contrary, to reduce labor mobility and more generally to promote some form of stability. Another novelty of the static framework explored above is to underscore the possible effects that technological, institutional, or policy-driven changes can have on cities. The urban equilibrium is determined by the interplay of the wage, cost of living, and labor supply curves. In turn, these curves are determined by a wide array of forces, which can all affect cities indirectly. Taking a more dynamic perspective does not fundamentally alter the recommendations of more static approaches.34 It leads us to put even more emphasis on the mobility of people and goods across places. This emphasis on mobility and flexibility in factor allocation and reallocation should also arguably be part of any modern growth agenda. Hence even though, at some fine level of detail, static and dynamic approaches to urban policy might conflict, these divergences are minor from a practical perspective. It is also important to note that an urban perspective on economic growth does not appear to conflict with any broader growth agenda. 34Practically, the distinction between "static" and "dynamic" is blurred. For instance, the construction of the U.S. interstate system should be viewed conceptually as a one-time improvement. Nonetheless, Fernald (1999) estimates that it generated about one point of annual GDP growth during nearly 20 years, arguably a long-run effect. Cities: Engines of Growth and Prosperity for Developing Countries? 43 This being said, implementing a broad-ranging urban agenda aimed at bolstering economic growth raises a number of problems. The first is that such an agenda is rather demanding since it includes raising the efficiency of public good provision, lowering barriers to mobility, improving market access to allow secondary cities to develop, and so forth. The second difficulty is that the political economy of many of these issues is often a formidable obstacle to change. Hence, politics and other more mundane feasibility constraints, such as the limited capabilities of many governments, require establishing priorities. On the other hand, the framework used above shows clearly that cities operate in a second-best world where fixing one problem may not result in any tangible improvement locally. Hence we face a policy dilemma where doing all at once may not be possible but a step-by-step approach may not be effective. Furthermore, growth agendas often identify a number of "growth drivers" that need to be fostered. Rather than drivers, it may be more fruitful in practice to think about constraints and bottlenecks to be removed. In this respect, the theoretical framework developed above can be useful to identify constraints to harmonious urban development.35 Since constraints and bottlenecks are likely to differ across countries, so will the diagnostic and, ultimately, the urban strategy. The main caveat with this diagnostic approach is that static constraints to urban development such as a grid-locked city are for all to see, while dynamic constraints are much more difficult to identify. The final question relates to who should be in charge of implementing any "cities and growth" agenda. The emphasis given here to the mobility of goods and factors between cities suggests that central governments should have a prominent role in promoting labor mobility, developing infrastructure, and removing impediments to internal trade. On the other hand, cities have also an important part to play to improve the life of their residents and minimize their cost of living. This division of labor between central and local governments is nonetheless unlikely to remain free of tensions. First, there is a fundamental asymmetry between primate cities and secondary cities. No secondary city can alone have an effect on the entire urban system whereas primate cities do. There is also considerable heterogeneity in the capabilities of secondary cities to design and implement local policies that would be consistent with a national growth agenda. 35Although this would go beyond the scope of this paper, the framework used above and its extensions could be further developed as a diagnostic tool in the spirit of Hausmann, Rodrik, and Velasco (2005). 44 Gilles Duranton References Ades, Alberto F. and Edward L. Glaeser. 1995. Trade and circuses: Explaining urban giants. Quarterly Journal of Economics 110(1):195­227. Agrawal, Ajay, Iain Cockburn, and John McHale. 2006. Gone but not forgotten: Knowledge flows, labor mobility, and enduring social relationships. Journal of Economic Geography 6(5):571­591. Alcalá, Francisco, and Antonio Ciccone. 2004. Trade and productivity. Quarterly Journal of Economics 119(2):613­646. Almeida, Paul, and Bruce Kogut. 1999. Localization of knowledge and the mobility of engineers in regional networks. Management Science 45(7):195­ 227. Amiti, Mary, and Lisa Cameron. 2007. Economic geography and wages. Review of Economics and Statistics 89(1):15­29. Aten, Bettina, and Alan Heston. 2005. Regional output differences in international perspective. In Ravi Kanbur and Anthony J. Venables, eds. Spatial Inequality and Development. New York: Oxford University Press. Au, Chun-Chung, and J. Vernon Henderson. 2006a. How migration restrictions limit agglomeration and productivity in China. Journal of Development Economics 80(2):350­388. ------. 2006b. Are Chinese cities too small? Review of Economic Studies 73(3):549­ 576. Baldwin, Richard E., Rikard Forslid, Philippe Martin, Gianmarco I. P. Ottaviano, and Frédéric Robert-Nicoud. 2004. Economic Geography and Public Policy. New Jersey: Princeton University Press. Barrios, Salvador, Luisito Bertinelli, and Eric Strobl. 2006. Climatic change and rural-urban migration: The case of sub-Saharan Africa. Journal of Urban Economics 60(3):357­371. Bartelsman, Eric, John Haltiwanger, and Stefano Scarpetta. 2004. Microeconomic evidence of creative destruction in industrial and developing countries. Processed, University of Maryland. Bertinelli, Luisito and Duncan Black. 2004. Urbanization and growth. Journal of Urban Economics 56(1):80­96. Bertinelli, Luisito and Eric Strobl. 2003. Urbanization, urban concentration and economic growth in developing countries. Research Paper 03/14, CREDIT, University of Nottingham. Cities: Engines of Growth and Prosperity for Developing Countries? 45 Black, Duncan and J. Vernon Henderson. 1999. A theory of urban growth. Journal of Political Economy 107(2):252­284. Brueckner, Jan K. 1990. Analyzing Third World urbanization: A model with empirical evidence. Economic Development and Cultural Change 38(3):587­ 610. Brueckner, Jan K. and Yves Zenou. 1999. Harris-Todaro models with a land market. Regional Science and Urban Economics 29(3):317­339. Combes, Pierre-Philippe, and Gilles Duranton. 2006. Labor pooling, labor poaching, and spatial clustering. Regional Science and Urban Economics 36(1):1­28. Combes, Pierre-Philippe, Gilles Duranton, and Henry G. Overman. 2005. Agglomeration and the adjustment of the spatial economy. Papers in Regional Science 84(3):311­349. Combes, Pierre-Philippe, Gilles Duranton, and Laurent Gobillon. 2008a. Spatial wage disparities: Sorting matters! Journal of Urban Economics 63(forthcoming). Combes, Pierre-Philippe, Gilles Duranton, Laurent Gobillon, and Sébastien Roux. 2008b. Estimating agglomeration economies with history, geology, and worker effects. University of Toronto. Combes, Pierre-Philippe, Thierry Mayer, and Jacques Thisse. 2008c. Economic Geography. Princeton (NJ): Princeton University Press. Forthcoming. Conley, Timothy G., Frederick Flyer, and Grace R. Tsiang. 2003. Spillover from local market human capital and spatial distribution of productivity in Malaysia. Advances in Economic Analysis and Policy 3(1):Article 5. da Mata, Daniel, Uwe Deichmann, J. Vernon Henderson, Somik V. Lall, and Hyoung Gun Wang. 2005. Examining the growth patterns of Brazilian cities. Policy Research Working Paper 3724, World Bank, Washington, D.C. ------. 2007. Determinants of city growth in Brazil. Journal of Urban Economics 61(forthcoming). Davis, James C., and J. Vernon Henderson. 2003. Evidence on the political economy of the urbanization process. Journal of Urban Economics 53(1):98­ 125. Davis, Steven J., John C. Haltiwanger, and Scott Schuh. 1996. Job Creation and Destruction. Cambridge, MA: MIT Press. De Soto, Hernando. 2000. The Mystery of Capital: Why Capitalism Triumphs in the West and Fails Everywhere Else. New York: Basic Books. 46 Gilles Duranton Deichmann, Uwe, Kai Kaiser, Somik V. Lall, and Zmarak Shalizi. 2005. Agglomeration, transport, and regional development in Indonesia. Policy Research Working Paper 3477, World Bank, Washington, D.C. Di Tella, Rafael, Sebastian Galliani, and Ernesto Schargrodsky. 2007. The formation of beliefs: Evidence from the allocation of land titles to squatters. Quarterly Journal of Economics 122(1):209­241. Durand-Lasserve, Alain, and Harris Selod. 2007. The formalisation of urban land tenure in developing countries. Processed, Paris School of Economics. Duranton, Gilles. 2006. Human capital externalities in cities: Identification and policy issues. In Richard J. Arnott and Daniel P. McMillen, eds., A Companion to Urban Economics. Oxford: Blackwell, 24­39. ------. 2007. Urban evolutions: The fast, the slow, and the still. American Economic Review 97(1):197­221. Duranton, Gilles and Diego Puga. 2001. Nursery cities: Urban diversity, process innovation, and the life cycle of products. American Economic Review 91(5):1454­1477. ------. 2004. Micro-foundations of urban agglomeration economies. In Vernon Henderson and Jacques-François Thisse, eds., Handbook of Regional and Urban Economics, volume 4. Amsterdam: North-Holland, 2063­2117. ------. 2005. From sectoral to functional urban specialization. Journal of Urban Economics 57(2):343­370. Durlauf, Steven N., Paul A. Johnson, and Jonathan W. Temple. 2005. Growth econometrics. In Philippe Aghion and Steven N. Durlauf, eds., Handbook of Economic Growth, volume 1A. Amsterdam: North-Holland, 555­677. Eaton, Jonathan, and Zvi Eckstein. 1997. Cities and growth: Theory and evidence from France and Japan. Regional Science and Urban Economics 27(4­5):443­ 474. Epple, Dennis, and Thomas Nechyba. 2004. Fiscal decentralization. In Vernon Henderson and Jacques-François Thisse, eds., Handbook of Regional and Urban Economics, volume 4. Amsterdam: North-Holland, 2423­2480. Fay, Marianne, and Charlotte Opal. 1999. Urbanization without growth: A not- so-uncommon phenomenon. Policy Research Working Paper 2412, World Bank, Washington, D.C. Fernald, John G. 1999. Roads to prosperity? Assessing the link between public capital and productivity. American Economic Review 89(3):619­638. Field, Erica. 2007. Entitled to work: Urban property rights and labor supply in Peru. Quarterly Journal of Economics 122(4):1561­1602. Cities: Engines of Growth and Prosperity for Developing Countries? 47 Foster, Lucia, John C. Haltiwanger, and C. J. Krizan. 2001. Aggregate productivity growth: Lessons from microeconomic evidence. In Charles R. Hulten, Edwin R. Dean, and Michael J. Harper, eds., New Developments in Productivity Analysis. Chicago: National Bureau of Economic Research and University of Chicago Press, 303­372. Franco, April M., and Darren Filson. 2006. Spin-outs: Knowledge diffusion through employee mobility. Rand Journal of Economics 37(4):841­860. Freedman, Matthew. 2007. Location decisions in a changing labor market environment. Processed, University of Maryland. Fujita, Masahisa, Paul R. Krugman, and Anthony J. Venables. 1999. The Spatial Economy: Cities, Regions, and International Trade. Cambridge, MA:MIT Press. Fujita, Masahisa, and Tomoya Mori. 1996.The role of ports in the making of major cities: Self-agglomeration and hub-effect. Journal of Development Economics 49(1):93­120. ------. 2005. Transport development and the evolution of economic geography. Portuguese Economic Journal 4(2):129­156. Glaeser, Edward L. 1999. Learning in cities. Journal of Urban Economics 46(2):254­ 277. Glaeser, Edward L., and David C. Maré. 2001. Cities and skills. Journal of Labor Economics 19(2):316­342. Greenwood, Michael J. 1997. Internal migrations in developed countries. In Mark R. Rosenzweig and Oded Stark, eds., Handbook of Population and Family Economics, volume 1B. Amsterdam: North-Holland, 647­720. Harris, Chauncy D. 1954. The market as a factor in the localization of industry in the United States. Annals of the Association of American Geographers 44(4):315­ 348. Harris, John R., and Michael P. Todaro. 1970. Migration, unemployment and development: A two-sector analysis. American Economic Review 60(1):126­ 142. Hausmann, Ricardo, Dani Rodrik, and Andrés Velasco. 2005. Growth diagnostics. Processed, Kennedy School of Government, Harvard University. Head, Keith, and Thierry Mayer. 2004. The empirics of agglomeration and trade. In Vernon Henderson and Jacques-François Thisse, eds., Handbook of Regional and Urban Economics, volume 4. Amsterdam: North-Holland, 2609­ 2669. ------. 2006. Regional wage and employment responses to market potential in the EU. Regional Science and Urban Economics 36(5):573­594. 48 Gilles Duranton Helsley, Robert W. 2004. Urban political economics. In Vernon Henderson and Jacques-François Thisse, eds., Handbook of Regional and Urban Economics, volume 4. Amsterdam: North-Holland, 2381­2421. Henderson, J. Vernon. 1974. The sizes and types of cities. American Economic Review 64(4):640­656. ------. 1988. Urban Development: Theory, Fact and Illusion. Oxford: Oxford University Press. ------. 1997. Medium size cities. Regional Science and Urban Economics 27(6):583­ 612. ------. 2002a. Urban primacy, external costs, and the quality of life. Resource and Energy Economics 24(1):95­106. ------. 2002b. Urbanization in developing countries. World Bank Research Observer 17(1):89­112. ------. 2003. The urbanization process and economic growth: The so-what question. Journal of Economic Growth 8(1):47­71. ------. 2005. Urbanization and growth. In Philippe Aghion and Steven N. Durlauf, eds., Handbook of Economic Growth, volume 1B. Amsterdam: North- Holland, 1543­1591. ------. 2007. Exclusion through informal sector housing development. Processed, Brown University. Henderson, J. Vernon, and Anthony J. Venables. 2006. The dynamics of city formation. Processed, Brown University. Henderson, J. Vernon, and Arindam Mitra. 1996. The new urban landscape: Developers and edge cities. Regional Science and Urban Economics 26(6):613­ 643. Henderson, J. Vernon, and Hyoung Gun Wang. 2007. Urbanization and city growth: The role of institutions. Regional Science and Urban Economics 37(3):283­313. Henderson, J. Vernon, and Randy Becker. 2000. Political economy of city sizes and formation. Journal of Urban Economics 48(3):453­484. Henderson, J. Vernon, John Quigley, and Edwin Lim. 2007. Urbanization in China: Policy issues and options. Processed, Brown University. Henderson, Vernon, Todd Lee, and Yung Joo Lee. 2001. Scale externalities in Korea. Journal of Urban Economics 49(3):479­504. Jacobs, Jane. 1969. The Economy of Cities. New York: Random House. Keller, Wolfgagng. 2004. International technology diffusion. Journal of Economic Literature 42(3):752­782. Cities: Engines of Growth and Prosperity for Developing Countries? 49 Krugman, Paul R. 1991. Increasing returns and economic geography. Journal of Political Economy 99(3):484­499. Krugman, Paul R., and Raúl Livas Elizondo. 1996. Trade policy and the Third World metropolis. Journal of Development Economics 49(1):137­150. Lall, Somik V., Harris Selod, and Zmarak Shalizi. 2006. Rural-urban migration in developing countries: A survey of theoretical predictions and empirical findings. Policy Research Working Paper 3915. World Bank, Washington, D.C. Lall, Somik V., Jun Koo, and Sanjoy Chakravorty. 2003. Diversity matters: The economic geography of industry location in India. Policy Research Working Paper 3072. World Bank, Washington, D.C. Lall, Somik V., Richard Funderburg, and Tito Yepes. 2004a. Location, concentration, and performance of economic activity in Brazil. Policy Research Working Paper 3268, World Bank, Washington, D.C. Lall, Somik V., Zmarak Shalizi, and Uwe Deichmann. 2004b. Agglomeration economies and productivity in Indian industry. Journal of Development Economics 73(3):643­673. Lucas, Robert E. Jr. 1988. On the mechanics of economic development. Journal of Monetary Economics 22(1):3­42. ------. 2004. Life earnings and rural-urban migration. Journal of Political Economy 112(1, part 2):S29­S59. Malpezzi, Stephen. 1999. Economic analysis of housing markets in developing and transition countries. In Edwin S. Mills and Paul Cheshire, eds., Handbook of Regional and Urban Economics, volume 3. Amsterdam: North- Holland, 1791­1864. Marshall, Alfred. 1890. Principles of Economics. London: Macmillan. Møen, Jarle. 2005. Is mobility of technical personnel a source of R&D spillovers? Journal of Labor Economics 23(1):81­114. Moomaw, Ronald L. and Ali M. Shatter. 1996. Urbanization and development: A bias towards large cities? Journal of Urban Economics 40(1):13­37. Moretti, Enrico. 2004. Human capital externalities in cities. In Vernon Henderson and Jacques-François Thisse, eds., Handbook of Regional and Urban Economics, volume 4. Amsterdam: North-Holland, 2243­2291. Nitsch, Volker. 2006. Trade openness and urban concentration: New evidence. Journal of Economic Integration 21(2):340­362. Overman, Henry G. and Anthony J. Venables. 2005. Cities in the developing world. Report, Department for International Development (UK). 50 Gilles Duranton Peri, Giovanni. 2002. Young workers, learning, and agglomerations. Journal of Urban Economics 52(3):582­607. Poelhekke, Steven. 2007. Urban growth, uninsured risk, and the rural origins of aggregate volatility. Unpublished, European University Institute. Ravallion, Martin and Quentin Wodon. 1999. Poor areas, or only poor people? Journal of Regional Science 39(4):689­711. Redding, Stephen, and Anthony J. Venables. 2004. Economic geography and international inequality. Journal of International Economics 62(1):63­82. Richardson, Harry W. 1987. The costs of urbanization: A four-country comparison. Economic Development and Cultural Change 35(3):561­580. Rosenthal, Stuart S. and William C. Strange. 2004. Evidence on the nature and sources of agglomeration economies. In Vernon Henderson and Jacques- François Thisse, eds., Handbook of Regional and Urban Economics, volume 4. Amsterdam: North-Holland, 2119­ 2171. Rossi-Hansberg, Esteban, and Mark L. J. Wright. 2007. Urban structure and growth. Review of Economic Studies 74(2):597­624. Saiz, Albert. 2006. Dictatorships and highways. Regional Science and Urban Economics 36(2):187­206. Solow, Robert M. 1956. A contribution to the theory of economic growth. Quarterly Journal of Economics 70(1):65­94. Thomas, Vinod. 1980. Spatial differences in the cost of living. Journal of Urban Economics 8(2):108­122. Timmins, Christopher. 2006. Estimating spatial differences in the Brazilian cost of living with households location choices. Journal of Development Economics 80(1):59­83. Wacziarg, Romain, and Karen Horn Welch. 2003. Trade liberalization and growth: New evidence. Working Paper 10152, National Bureau of Economic Research. Wheeler, Christopher H. 2006. Cities and the growth of wages among young workers: Evidence from the NLSY. Journal of Urban Economics 60(2):162­184. Cities: Engines of Growth and Prosperity for Developing Countries? 51 Eco-Audit Environmental Benefits Statement The Commission on Growth and Development is committed to preserving endangered forests and natural resources. The World Bank's Office of the Publisher has chosen to print these Working Papers on 100 percent postconsumer recycled paper, processed chlorine free, in accordance with the recommended standards for paper usage set by Green Press Initiative--a nonprofit program supporting publishers in using fiber that is not sourced from endangered forests. For more information, visit www.greenpressinitiative.org. The printing of all the Working Papers in this Series on recycled paper saved the following: Trees* Solid Waste Water Net Greenhouse Gases Total Energy 48 2,247 17,500 4,216 33 mil. *40 inches in height and 6­8 Pounds Gallons Pounds CO2 Equivalent BTUs inches in diameter The Commission on Growth and Development Working Paper Series 3. Normalizing Industrial Policy, by Dani Rodrik, March 2008 4. The Real Exchange Rate and Economic Growth, by Barry Eichengreen, March 2008 5. Globalization, Growth and Distribution: Framing the Questions, by Ravi Kanbur, March 2008 6. Growth Strategies and Dynamics: Insights from Country Experiences, by Mohamed A. El-Erian and Michael Spence, March 2008 7. Political Competition, Policy Making, and the Quality of Public Policies in Costa Rica, by Fabrice Lehoucq, April 2008 8. Leadership for Growth, Development, and Poverty Reduction: An African Viewpoint and Experience, by Benjamin William Mkapa, April 2008 9. Growth Strategies for Africa, by Paul Collier, April 2008 10. The Role of Institutions in Growth and Development, by Daron Acemoglu and James Robinson, April 2008 11. Finance, Financial Sector Policies, and Long-Run Growth, by Asli Demirgüç-Kunt and Ross Levine, April 2008 12. Cities: Engines of Growth and Prosperity for Developing Countries? by Gilles Duranton, April 2008 Forthcoming Papers in the Series: Housing Policy in Developing Countries: The Importance of the Informal Economy, by Richard Arnott (April 2008) Economic Growth in Egypt: Impediments and Constraints (1974­2004), by Hazem El Beblawi (April 2008) Electronic copies of the working papers in this series are available online at www.growthcommission.org. They can also be requested by sending an e-mail to contactinfo@growthcommission.org. T his paper reviews the evidence about the effects of urbanization and cities on productivity and economic growth in developing countries using a consistent theoretical framework. Just like in developed economies, there is strong evidence Commission on Growth and Development that cities in developing countries bolster productive efficiency. Regarding wheth- Montek Ahluwalia Edmar Bacha er cities promote self-sustained growth, the evidence is suggestive but ultimately Dr. Boediono inconclusive. These findings imply that the traditional agenda of aiming to raise Lord John Browne within-city efficiency should be continued. Furthermore, reducing the obstacles Kemal Dervis ¸ to the reallocation of factors across cities is also desirable. Alejandro Foxley Goh Chok Tong Gilles Duranton, Noranda Chair in International Trade and Development in Han Duck-soo the Department of Economics, University of Toronto Danuta Hübner Carin Jämtin Pedro-Pablo Kuczynski Danny Leipziger, Vice Chair Trevor Manuel Mahmoud Mohieldin Ngozi N. Okonjo-Iweala Robert Rubin Robert Solow Michael Spence, Chair Sir K. Dwight Venner Ernesto Zedillo Zhou Xiaochuan The mandate of the Commission on Growth and Development is to gather the best understanding there is about the policies and strategies that underlie rapid economic growth and poverty reduction. The Commission's audience is the leaders of developing countries. The Commission is supported by the governments of Australia, Sweden, the Netherlands, and United Kingdom, The William and Flora Hewlett Foundation, and The World Bank Group. www.growthcommission.org contactinfo@growthcommission.org