The Speed of Urbanization and Economic Development:A Comparison of Industrial Europe and Contemporary Africa∗

Luc CHRISTIAENSENa Marina GINDELSKYb Rémi JEDWABc

a The World Bankb Department of Economics, George Washington Universityc Department of Economics, George Washington University

This Version: May 10th, 2013

Preliminary Draft - Please do not cite or quote without permission

Abstract: Africa’s level of urbanization is relatively high for its level of eco-nomic development. Standard explanations for this paradox include the urbanbias and rural poverty. In the Harris-Todaro model, there is rural-to-urban mi-gration as long as the expected urban wage is higher than the rural wage.Many models explain the push and pull factors of rural-to-urban migration.Land shortages and natural disasters constitute common rural push factors,while agricultural taxation and public employment are examples of urban pullfactors. In this paper, we offer an alternative explanation based on an ur-ban push. African cities can be classified as “mushroom cities” vs. the “killercities” of Industrial Europe; Fertility remains relatively high, while mortalityhas fallen to relatively low levels, due to colonial investments and the epi-demiological transition. Natural increase has become the main source of ur-banization, causing the population of African cities to double every 15 years,compared with 35 years in Industrial Europe, with other developing regionslying in between. We construct a novel data set on the urban demographictransition for 7 European countries over 100 years (1800-1910) and 32 de-veloping countries over 50 years (1960-2010). We then discuss the welfareimplications of an increased speed of urbanization. In line with Malthusianmodels of economic growth, we provide qualitative evidence that these laborsupply shocks led to congestion effects and an “urbanization” of poverty.

Keywords: Urbanization; Demographic Transition; Urban Poverty; AfricaJEL classification: J11; J21; N97; O18; R11; R23

∗Luc Christiaensen, The World Bank (e-mail: lchristiaensen@worldbank.org). Marina Gindelsky, George Washington Univer-sity (e-mail: marina_g@gwu.edu). Remi Jedwab, George Washington University (jedwab@gwu.edu). We would like to thankPaul Carrillo, Carmel Chiswick, Jeremiah Dittmar, James Foster, Fabian Lange, John Nye, Stephen Smith, Jan de Vries, JeffreyWilliamson, Tony Yezer and seminar participants at GWU for very helpful comments. We gratefully acknowledge the support ofthe Institute for International Economic Policy at GWU.

1. INTRODUCTION

Sub-Saharan Africa has experienced dramatic urban growth over the past fifty years. The number of urbanresidents has grown an unprecedented 5% annually after independence (1960-2010), as compared withless than 2% annually for European countries during the Industrial Revolution (1800-1910). During theseperiods, the urbanization rate of both Europe and Africa increased from 10% to 40%. Therefore, Africancountries have experienced the same growth in urbanization as Europe in half the time.

While urban growth was associated with industrialization in Europe, Africa’s level of urbanization is rela-tively high for its level of economic development. For example, Africa is as urbanized as Asia, although itis three times poorer. Standard explanations for this paradox include rural poverty - (a rural push) and theurban bias (an urban pull). Land shortages and natural disasters constitute common rural push factors,while agricultural taxation and (urban-based) public employment are examples of urban pull factors. Inthis paper, we offer an alternative explanation based on an urban push.

Retrospectively, the cities of Industrial Europe can be classified as “killer cities”. Mortality was high, dueto industrial pollution, the lack of sanitation, and inadequate health facilities while fertility was relativelylow. Consequently, in these European cities, rural-to-urban migration was necessary for growth in orderto compensate for the overall negative, or neutral, rate of natural increase. European cities could not havegrown without a massive influx of rural migrants striving for a better life.

Conversely, birth rates in African cities remain high, while mortality has fallen to relatively low levels, dueto colonial investments and the epidemiological transition. The growth of African cities was thus fueled byboth migration and natural increase. Given the large contribution of natural increase, urbanization becamea fast-paced self-reinforcing process, allowing us to classify these African cities as "mushroom cities". Usinghistorical data on the urban demographic transition for Industrial Europe and contemporary Sub-SaharanAfrica and a new conceptual framework, we show that the two urbanization processes differed in severaldimensions: speed (fast vs. even faster), type (migration vs. natural increase) and relationship witheconomic growth (urban wealth vs. urban poverty).

First, we provide historical evidence on the rapid growth of African cities, relative to Industrial Europe inthe 19th century and other developing regions after World War II. We find that the average growth rateof the urban population has been 4.7% a year in post-1950 Africa, vs. 2.0% a year in Industrial Europein 1800-1910, 3.2% in Latin America and the Caribbean (LAC), 3.4% in Asia and 4.0% in the Middle-Eastand North Africa (MENA) post-1950.1

Second, we show that the rapid growth of African cities was mostly driven by natural increase, i.e. anurban push, and not by migration as in Industrial Europe. We collect historical data on the crude ratesof birth and death for urban areas and rural areas of 7 European (or Neo-European) countries in the19th century (1800-1910) and 10 African countries post-1950 (1950-2010). We confirm that the citiesof Industrial Europe were “killer cities”, where mortality was high and fertility was low. The contributionof natural increase to urban growth was around 0.5% a year. Given a speed of urbanization of 2.0%, theaverage contribution of migration was around 1.5%. On the contrary, the cities of contemporary Africa are“mushroom cities”, where fertility is high and mortality is low. The contribution of urban natural increasewas around 2.9% a year. Given a speed of 4.7%, the average contribution of migration was around 1.8%.This difference in the rate of natural increase causes the population of African cities to double every 15years, compared with 35 years in Industrial Europe. We also collect data for 22 non-African developingcountries post-1950 and find that the average contribution of urban natural increase was 2.6% in theMENA region, 2.2% in the LAC region and 1.6% in Asia. This contribution decreased almost everywhere,as a result of the fertility transition. Yet it remained high in Africa, at around 2.6% post-2000. Africancities will likely keep growing for several decades, until they complete their fertility transition.

Third, we develop a conceptual framework that allows us to explain the potential implications of the

1As these different speeds of urbanization could reflect continental differences in the urban definition adopted by each groupof countries, we verify that this result is robust to using other measures of urbanization for African countries and excludingfrom the analysis a few specific countries that have been identified as problematical in the urban studies literature. We also usehistorical data on the growth of the largest city of each country, instead of the whole urban population, and find similar results.

2

speed of urbanization and the different types of urban growth – migration vs. natural increase – onurban incomes. In our theoretical exercise, there are two types of urbanization. In the first economy,urbanization is associated with migration and rising urban incomes. In the first period, if the urban-ruralwage differential is positive, rural residents migrate to the cities and the urban population increases. Inthe second period, if agglomeration effects dominate congestion effects, urban productivity and wagesincrease further, which attracts even more migrants, and so on in the following periods. This narrativecould describe well the historical experience of Industrial Europe, but could also fit the patterns recentlyfound in Asian countries. In the second economy, urbanization is associated with natural increase anddeclining urban incomes, because the speed of urbanization is too fast. In the first period, if the urban-rural wage differential is positive, rural residents migrate to the cities and the urban population increases.In the second period, urban labor supply shocks mechanically increase the size of the urban population,but urban productivity and wages decrease if congestion effects dominate agglomeration effects, and so onin the following periods. This narrative could suit the recent patterns found in African countries. We thenprovide qualitative evidence on congestion effects in African cities and hypothesize that urban povertyincreased and migration rates decreased as a result. Lastly, we discuss the role of public policy in ensuringthat urban growth and economic growth go hand in hand.

We focus on Africa for two reasons. First, it is an extreme example of an urban push. Industrial Europewas an extreme example of a rural push and an urban pull. Post-1950, the speed of urbanization wasrelatively faster in Africa than in Asia, LAC and MENA. Second, many African countries are also examplesof overurbanization. There are large urban populations and agglomerations in even the poorest Africancities (e.g., Kinshasa in the Democratic Republic of the Congo or Addis Ababa in Ethiopia), and the questionis how these cities grow. Moreover, urban growth in Asia more closely followed to the European model, asit was associated with migration and rising urban incomes.

This paper is based on a large body of work on the relationship between urbanization and economicgrowth. In the Harris-Todaro model, there is rural-to-urban migration as long as the (expected) urbanwage is higher than the rural wage. There is urbanization with growth if urban growth is associated withrising per capita incomes. There is urbanization without growth if a country’s level of urbanization isrelatively high for its level of economic development (Bairoch, 1988; Fay & Opal, 2000).

The structural change literature associates urbanization with growth, structural change being understoodas the decline in (rural-based) agriculture and the rise of (urban-based) manufacturing and services. The“labor push” approach shows how a rise in agricultural productivity (what we might think of as a Green Rev-olution) reduces the “food problem” and releases labor for the modern sector (Schultz, 1953; Matsuyama,1992; Caselli & Coleman II, 2001; Gollin, Parente & Rogerson, 2002; Voigtländer & Voth, 2006; Nunn& Qian, 2011; Michaels, Rauch & Redding, 2012). The “labor pull” approach describes how an increasein non-agricultural productivity (an industrial revolution) attracts underemployed labor from agricultureinto the modern (manufacturing) sector (Lewis, 1954; Harris & Todaro, 1970; Hansen & Prescott, 2002;Lucas, 2004; Alvarez-Cuadrado & Poschke, 2011). As shown by Gollin, Jedwab & Vollrath (2013) andJedwab (2013), it is quite possible for an economy to urbanize without any changes in agricultural andmanufacturing productivity, if the country experiences a natural resource revolution. If resource windfallsare spent on urban goods and services, the country can urbanize owing to the emergence of “consumptioncities”. This theory is in line with the “labor pull” approach: however, it considers the consumption ofresource rents as the main driver of the urban-rural income gap. While urbanization is usually associ-ated with growth, the green revolution, the industrial revolution and the natural resource revolution haveplayed various roles in the urbanization process. For example, the most urbanized countries in Asia exportmanufactured goods, while the most urbanized countries in Africa export natural resources.

However, a few studies argue that some parts of the world (Sub-Saharan Africa in particular) have ur-banized without high economic growth (Bairoch, 1988; Fay & Opal, 2000). This excessive urbanization isoften attributed to the urban bias, i.e. urban-biased policies such as agricultural taxation, (urban-based)public employment and food price subsidies. These urban pull factors led to overurbanization and primacyin developing countries (Lipton, 1977; Bates, 1981; Ades & Glaeser, 1995; Davis & Henderson, 2003). Ru-ral poverty, land pressure and natural disasters (e.g., climate change) also constituted rural push factors

3

feeding rural exodus (Barrios, Bertinelli & Strobl, 2006; Poelhekke, 2010). Yet, commodity booms couldalso explain why Africa is relatively urbanized for its level of economic development. If the African coun-tries that urbanized as a result of commodity booms in the 1960s-1970s experienced economic declinewith the debt crisis of the 1980s-1990s, and if there are fixed costs of building cities, African cities mayhave persisted even in the face of declining incomes (Jedwab, 2013).

Our paper builds on the previous literature in three important ways. First, we show that natural increasehas been the main factor of urban growth in Africa post-1960. The Harris-Todaro model ill-suits our pur-pose, since urban growth comes from exogenous labor supply shocks. Second, while there has been dis-cussion of urban pull and rural push factors in the literature, we show that urban push factors should alsobe considered, especially in Africa where natural increase was high. Third, we reconcile the two literatureson urbanization with and without growth. The urbanizing effects of high migration rates in the 1960s,due to urbanization with growth, were amplified by the urban demographic transition. If this populationgrowth decreases urban incomes, urban growth is now disconnected from economic growth.

Since urbanization is often considered to be good for long-run growth, it is important to study this topic.The economic geography literature suggests that agglomeration promotes growth, in both developed coun-tries (Rosenthal & Strange, 2004; Henderson, 2005; Glaeser & Gottlieb, 2009; Greenstone, Hornbeck& Moretti, 2010) and developing countries (Overman & Venables, 2005; Henderson, 2010; Felkner &Townsend, 2011). Given that urbanization is a form of agglomeration, it has been argued that cities couldpromote growth in developing countries (Duranton, 2008; World Bank, 2009; Venables, 2010; McKin-sey, 2011).2 What we hypothesize is that higher urban densities could lead to multiple equilibria, urbanwealth or urban poverty, depending on the respective contributions of agglomeration forces and conges-tion effects, which depend on the speed of urbanization. African cities could be associated with increasingpoverty, due to congestion. The urban push could then be a factor of the “urbanization of global poverty”(Ravallion, Chen & Sangraula, 2007).

This paper also contributes to the literature on the relationship between demographic growth and eco-nomic growth. Theoretically, a rise in (urban) population density could promote growth, if the numberof inventors increases - spurring technological change, if returns-to-scale in infrastructure promote humancapital accumulation or if there are agglomeration economies (Kremer, 1993; Becker, Glaeser & Murphy,1999; Lagerlöf, 2003). In the Malthusian model, however, population growth has a negative effect on percapita income because land is inelastically supplied and the land-labor ratio decreases. Similarly, in theSolow model, population growth decreases the stock of physical capital per person, which then reducesper capita income. In the standard Malthusian model of economic growth, any positive income shock istemporary; fertility increases and mortality decreases, so that any increase in total income is negated. Inother words, the stock of capital per capita is stable (and low) in the long-run.3 Countries can only developin the long-run if technology progresses and the demographic transition limits population growth (Weil &Galor, 1999, 2000; Hansen & Prescott, 2002; Lagerlöf, 2006). If the economy is Malthusian, any exoge-nous increase in population should decrease the capital-labor ratio and per capita income. Symmetrically,any exogenous decrease in population should increase the capital-labor ratio and per capita income.

A few studies have examined the effects of disease eradication campaigns on mortality, population growthand economic development. Acemoglu & Johnson (2007) show how the epidemiological transition of themid 20th century (e.g., the determination of the chemical structure and the mass production of penicillinfrom 1945) resulted in widespread and significant declines in mortality. As life expectancy rose, populationboomed in the poorer and more disease-prone countries. Though total GDP did not change much, incomeper capita decreased. However, the effects of health improvement on GDP per capita are potentially high

2For instance, McKinsey (2011) writes (p.3-19): “Africa’s long-term growth also will increasingly reflect interrelated socialand demographic trends that are creating new engines of domestic growth. Chief among these are urbanization and the rise ofthe middle-class African consumer. [...] In many African countries, urbanization is boosting productivity (which rises as workersmove from agricultural work into urban jobs), demand and investment.”

3During the “Malthusian” regime of economic growth, the most advanced societies have larger populations, but not signifi-cantly higher incomes (Diamond, 1997; Ashraf & Galor, 2011). Vollrath (2011) shows how high labor intensity in agriculture (asin rice production in Asia) led not only to higher population densities, but also to lower income per capita and a larger share oflabor in agriculture.

4

(Weil, 2007). If disease eradication campaigns also increase the “quality” of the labor force (e.g., healthierpeople are more educated and better workers), the positive effect of human capital could dominate thenegative effect of reduced mortality. Bleakley (2007), Bleakley & Lange (2009) and Bleakley (2010) allfind that disease eradication in the Americas led to an increase in both population and income per capita,mostly as a result of the human capital channel. On the contrary, Cutler et al. (2010) find that diseaseeradication had no effect on human capital and only a modest positive effect on income per capita in SouthAsia. The results of health interventions are thus sensitive to the context, depending on how Malthusianthe economy is and whether educational achievement increases alongside life expectancy.

Other studies have looked at the effects of decreases in population on economic development, whetherthese are caused by disease, war or fertility restrictions. Young (2005) shows how the AIDS epidemicmay raise future living standards in South Africa. If the negative effect of the epidemic on human capitalaccumulation is dominated by a decrease in fertility, and if the economy is Malthusian, it may increaseper capita income for the “survivors”. Similarly, the Black Death of 1348-1350 permitted an initial rise inEuropean per capita incomes that led to a transition to self-sustaining growth – the Industrial Revolution –a few centuries later. Voigtländer & Voth (2009) and Voigtländer & Voth (2013b) show how the initial risein income exacerbated, rather than reduced, mortality, as a result of three channels: (i) urbanization: asthe demand for urban products increased, urban centers grew in cities, but the congestion kept death rateshigh, (ii) trade: the demand for foreign products spread disease, and (iii) wars: higher tax revenues fi-nanced war between European powers. Voigtländer & Voth (2013a) also attribute the “European MarriagePattern”, the fact that women would marry late and have fewer children in Western Europe (Hajnal, 1965),to the Black Death; land abundance resulting from the epidemic triggered a shift towards land-intensivepastoral agriculture, improving female employment prospects and delaying the marriage age of women.In other words, the Black Death created a shift in the birth and death rates that constrained populationgrowth, while per capita incomes were slowly rising, until the onset of the Industrial Revolution. Ashraf,Weil & Wilde (2011) use a general equilibrium model simulated with Nigerian data to show that reducingfertility rates would increase per capita income in Africa through lower dependency rates, physical andhuman capital accumulation and less resource congestion.4

In our paper, we use an exogenous increase in population, studying it from the perspective of cities, notcountries. Cities are centers of human capital accumulation and agglomeration economies, so popula-tion growth could increase productivity (Becker, Glaeser & Murphy, 1999). However, since urban spaceis constrained, the potential for congestion effects is also much higher. The net effect of urban popula-tion growth on urban income is therefore ambiguous. If African cities are “Malthusian”, the congestioneffects dominate, and these cities pauperize. The two closest papers to our analysis are Young (2005) andVoigtländer & Voth (2013b). In Young (2005), any decrease in Africa’s population increases per capita in-come, while in our conceptual framework, any increase in Africa’s population decreases per capita income.In Voigtländer & Voth (2013b), any increase in urban mortality increases per capita income, while in ourconceptual framework, any decrease in urban mortality decreases per capita income.

The paper is organized as follows: Section 2 provides historical evidence of the rapid growth of Africancities post-World War II. Section 3 shows that the fast speed of urbanization in Africa can be explained bynatural increase, i.e. an urban push, while section 4 develops a conceptual framework that allows us todiscuss the welfare implications of this rapid urban growth. Section 5 concludes.

2. THE SPEED OF URBANIZATION: HISTORICAL EVIDENCE

In this section we provide historical evidence on the rapid growth of African cities, relative to IndustrialEurope in the 19th century and other developing regions after World War II.

4On the contrary, Lagerlöf (2003) explains that the escape from the Malthusian trap is the result of a series of mild epidemicshocks. While severe epidemic shocks constrain human capital accumulation, mild shocks are necessary for any improvement inthe technological environment.

5

2.1 Urbanization Rates and Economic Development

Urbanization is strongly associated with economic development (Acemoglu, Johnson & Robinson, 2005;Henderson, 2005, 2010). Countries were relatively unurbanized before the Industrial Revolution. Forexample, the most advanced civilizations before the 19th century had urbanization rates of around 10%-15%, as seen in Italy, Spain, China or India. When a few countries industrialized and developed, theirurbanization rate dramatically increased, usually from 10% to 40% (Bairoch, 1988). Figure 1 shows theurbanization rate, i.e. the share of the urban population in the total population, for Industrial Europefrom 1700-1950. In our analysis, Industrial Europe includes Western European countries and one Neo-European country, the United States. The urbanization rate was stable (around 12.5%) until 1800 andincreased to 41.3% in 1910. Countries that industrialized earlier also urbanized earlier. For example, theUnited Kingdom reached the 40% threshold in 1850, while Germany, France and the United States did notreach this threshold before the late 19th century.

Figure 2 shows the urbanization rate for four developing regions. The LAC region had already passed the40% threshold in 1950, while the MENA region did not pass it until 1970. In 1950, Africa and Asia wereboth poor and unurbanized, with an urbanization rate of around 10-15% (11.0% in Africa vs. 16.7% inAsia). In 2010, their urbanization rate was around 40% (36.6% in Africa vs. 42.5% in Asia). Since Africais three times poorer than Asia, it is relatively urbanized for its level of economic development. In ouranalysis, we focus on the 1800-1910 period for Europe and the 1950-2010 period for Africa. During theseperiods, the urbanization rates of both Europe and Africa increased from 10% to 40%. Therefore, Africancountries have experienced the same growth in urbanization as Europe in half the time.

What is the relationship between the urbanization rate and urban growth, i.e. the speed of urbanization?The urbanization rate U is the ratio of the urban population U POP to the total population POP, which isthe sum of the rural population RPOP and the urban population U POP:

U =U POP

POP=

U POP

U POP + RPOP=

1

1+ RPOP/U POP(1)

The urbanization rate only increases if the urban growth rate is higher than the rural growth rate:

4U > 0 ⇔ 4U POP >4RPOP (2)

Urban growth is not sufficient to increase the urbanization rate, but urban growth and the change in theurbanization rate are likely to be positively correlated. Urban growth is also interesting to study in andof itself. For example, urban congestion comes from rising urban densities, i.e. an increase in the urbanpopulation for a given urban zone, no matter the evolution of the national urbanization rate. In the restof the analysis, we will mostly study the urban growth rate, i.e. the speed of urbanization.

2.2 Urban Growth Rates in Industrial Europe

Figure 3 shows the urban growth rate for Industrial Europe from 1700-1950. While it was relatively lowthrough the early 18th century, it increased after 1750 and peaked in the late 19th century, at 2.2% peryear. It then declined in the 20th century, as population growth rates also slowed down and urbanizationrates converged towards 100%. In the 1800-1910 period, the urban growth rate was 2.0% per year.However, these rates hide significant differences across countries.

Urban growth rates accelerated for countries that industrialized and whose cities received a massive influxof migrants. As the industrialization of Europe and North America was sequential, so was the evolutionof urban growth rates. The urban growth rate accelerated as soon as the late 18th century in the UnitedKingdom. It accelerated in Germany, Belgium and France in the first half of the 19th century, beforeaccelerating everywhere else late in the 19th century. The United States is a clear exception to this pattern.It did not industrialize before 1850, yet the urban growth rate was very high, at 4.2% per year. In the 18th

6

century, there were very few cities, and those were very small. The early waves of international migrantsled to a boom of these cities and the urban growth rate accelerated. After 1850, it remained high, butdecelerated in the 20th century, as in Europe.

The urban growth rate of Industrial Europe was almost the same when we consider the largest city of eachcountry instead (1.8% against 2.0% per year during the 1800-1910 period). In other words, the largestcities did not grow faster than the rest of the urban population on average. Yet some individual cities grewrelatively faster during the period. Table 1 displays the growth rates of selected cities in Europe and theUnited States. The fastest-growing cities were on the U.S. East Coast (e.g., New York, Boston, Philadelphiaand Baltimore) and in the industrial regions of the United Kingdom (e.g., Manchester, Belfast, Glasgowand Liverpool). City growth was slower in the rest of Europe, particularly Southern Europe.

2.3 Urban Growth Rates in Africa and the Developing World

Figure 4 shows the urban growth rate for the four developing regions from 1960-2010. We find that theaverage growth rate of the urban population has been 4.7% a year in Africa post-1950, compared to 3.4%,3.2% and 4.0 in Asia and the LAC and MENA regions respectively. Urban growth was thus relatively fasterin Africa than in the rest of the developing world, and even more so than in Industrial Europe. A rate of4.7% implies that the urban population doubles every 15 years, while a rate of 2.0% as seen in IndustrialEurope means that population doubles every 35 years. It seems these rates peaked in the 1950s or 1960s.They have been declining since, although they are is still high in Africa today, at 3.8% per year.

Figure 5 displays the urban growth rates for each Sub-Saharan African country. These rates were highin countries that dramatically urbanized between 1960 and 2010. Most of these countries urbanized dueto natural resource exports (Jedwab, 2013; Gollin, Jedwab & Vollrath, 2013). For example, Botswanaurbanized after the discovery of diamonds in the late 1960s; the Ivory Coast urbanized as a result ofthe cocoa boom in the 1960s-1970s; and Angola urbanized after the discovery of oil in the 1970s. Afew countries also urbanized due to natural disasters, such as Mauritania after repeated droughts in the1970s-1980s (Barrios, Bertinelli & Strobl, 2006). Urban growth rates were then lower in countries thatwere already somewhat urbanized at independence, such as Ghana, the Republic of the Congo, the CentralAfrican Republic and South Africa. Lastly, urban growth rates were relatively lower in the poorer war-torncountries, such as Sierra Leone and the Democratic Republic of the Congo. While noteworthy, analysis ofthe variations at the country level within Africa is beyond the scope of this paper.

Though there is a concern that different speeds of urbanization could reflect continental differences in theurban definition adopted by each group of countries (Satterthwaite, 2007), we do not believe this is thecase. For example, countries sometimes adopt an administrative definition, according to which cities areadministrative centers of territorial units (e.g., provinces, districts, communes, etc.). Alternatively citiescould be classified as localities whose population is above a threshold of X inhabitants (e.g., 10,000, 5,000or 2,500). The urban growth rate for one group of countries could be overestimated or underestimatedrelative to the rest of the world, if the definition chosen for urbanization is significantly different for thisspecific group. Potts (2012) shows that the urbanization rate could be overestimated for a few countriesfor which the quality of the data is poor. If these countries represents a large share of Sub-Saharan Africa’spopulation, these definitional issues could affect the measurement of Africa’s urban growth.

First, considering the largest city only, we obtain similar urban growth rates as for the whole urban popula-tion. Indeed, the measurement of the population size of the largest city is not affected by these definitionalissues. Figure 6 shows the growth rate of the largest city for the four developing regions. Magnitudes arerelatively similar to those shown in Figure 4 (whole population). The speed of urbanization was evenfaster for Sub-Saharan Africa, at 4.9% per year on average vs. 2.5%, 2.6% and 3.1% in Asia and theLAC and MENA regions respectively. Table 2 displays the growth rates for selected African cities from1950-2010. These growth rates are also unlikely to be affected by the urban definition. While the averagegrowth rate was 5.5% per year, it was even higher for many cities (e.g., Yaoundé in Cameroon, Abidjanin Ivory Coast or Kigali in Rwanda). Interestingly, the growth rate was lower for South African cities, as aresult of the Apartheid policy that constrained rural-to-urban migration.

7

Second, we verify that these results are robust to other measures of urbanization for African countries.The main source of our data is World Urbanization Prospects (United Nations, 2011), which has beencriticized by the urban studies literature. Another data set is Geopolis (2012), an attempt by geographersto reconstruct the urban population of all countries using original census data and consistent measures ofurbanization.5 They have data for 35 countries in West Africa, Central Africa or Eastern Africa. SouthernAfrican countries are still missing from their analysis. Using these countries only, the urban growth ratewas 5.0% in the UN sample, vs. 4.7% in the full sample. Using Geopolis data, the urban growth rate was5.5%, which is even higher than our estimate. All in all, we are confident that the speed of urbanizationhas been relatively higher in Africa than in the rest of the world, across space and time.

3. THE URBAN PUSH AND THE SPEED OF URBANIZATION

In this section we show that the rapid growth of African cities was mostly driven by natural increase, i.e.an urban push, and not by rural-to-urban migration as in Industrial Europe.

3.1 Urban Pull, Rural Push and Urban Push

Urban growth can be decomposed into four components: natural increase, rural-to-urban migration,international-to-urban migration and urban reclassification. These are the equations of urban growth:

U POPt − U POPt−1 = [UCBRt−1− UC DRt−1] ∗ U POPt−1+ RM IGt + I M IGt + URECt (3)

U POPt − U POPt−1 = UCRN It−1 ∗ U POPt−1+ RM IGt + I M IGt + URECt (4)

where U POPt is the urban population at time t, UCBRt−1 and UC DRt−1 are the urban crude rates ofbirth and death at time t-1, while UCRN It−1 is the urban crude rate of natural increase, i.e. the differencebetween the two, RM IGt is the number of net rural-to-urban migrants between times t-1 and t, I M IGtis the number of net international-to-urban migrants between times t-1 and t, and URECt is the numberof rural residents reclassified as urban between times t-1 and t. Equations (3) and (4) must be dividedby the urban population at time t-1 (U POPt−1) to be expressed in percentage form. We now discuss therespective roles of the four potential components of urban growth:

• Urban natural increase: The urban crude rate of natural increase (UCRNI) is the urban crude birthrate (UCBR) minus the urban crude death rate (UCDR). If fertility is high and mortality is low, thenatural increase rate is positive. If it is positive, the urban population increases as a result of an urbanpush. While it is well-known that natural increase rates are high in developing countries, especiallyduring the second stage of the demographic transition, little is known about the differentials betweenthe urban and the rural natural increase rates. Here, urban growth is a direct function of the urbancrude natural increase rate (UCRNI), which must be estimated. The rural rate of natural increasemay affect urban growth through rural-to-urban migration or reclassification, as discussed below.

• Rural-to-urban migration: In the Harris-Todaro model, there is rural-to-urban migration (RM IGt >

0) as long as the (expected) urban real wage is higher than the rural real wage. However, asin the Rosen-Roback model, we must compare utilities and not wages, since preferences could beendogenous. This wage (or utility) differential could be the result of an urban pull or a rural push.For example, if the country experiences an industrial revolution or a natural resource revolution, theurban wage increases relatively more, which attracts workers from the countryside (urban pull). Ifthe government adopts urban-biased policies, the urban wage will increase and people will migrateto cities (urban pull). Rural poverty due to land pressure or (rural) natural disasters relativelydecreases the rural wage, causing rural migrants to flock to cities (rural push). A high rural rate ofnatural increase (RCRNI) would contribute to growing land pressure, if land supply is fixed (or ifthe costs of land colonization are high).

5In their analysis, they classify cities as all the localities that are likely to have more than 10,000 inhabitants in 2020, usingextrapolation. They then reconstruct the population history of each city from 1960 to date, using mostly census reports.

8

• International-to-urban migration: In the Harris-Todaro model, there is international-to-urban migra-tion (I M IGt) as long as the (expected) urban real wage is higher than the real wage in the countryof origin. The growth of American cities has been fueled by international migration for two centuriesnow, as exemplified by the successive waves of immigration from Europe in the 19th century or theGreat Mexican Emigration of the 20th century. On the contrary„ international-to-urban migrationdid not contribute much to urban growth in Industrial Europe. Migration rates were even negativefor a few countries. Had these emigrants been from cities and not left, the urban growth rates wouldhave been higher in their home countries. Net migration rates were also negative for most Africancountries. They were quite low during the 1960-2010 period, around -0.03% on average. Althoughthese rates varied across countries, they probably did not alter the urban growth rates much.

• Urban reclassification: “Rural” land is sometimes reclassified as “urban” for arbitrary reasons, e.g.when a rural suburb is absorbed by an agglomeration, or when a locality becomes “urban” given theurban definition. For example, in many countries, a locality is considered urban if its population sizeexceeds a certain population threshold, e.g. 2,000 or 5,000 inhabitants. Any rural locality that growsabove the threshold as a result of (rural) natural increase or (rural-to-rural) migration is now urban.In other countries, district capitals or “chef-lieux de communes” are considered cities. With politicaldecentralization, the number of administrative cities increases and this contributes to urban growth,although it is a statistical construction. This type of urban growth is purely definitional, although itcould still be considered as “real” urban growth.

Our goal is not to estimate the respective contributions of each component, but to isolate the effect ofthe urban rate of natural increase, i.e. the urban push. In the following sections, we discuss two types ofcities. The cities of Industrial Europe were "killer cities": mortality was high, due to industrial pollution,lack of sanitation, and inadequate health facilities. Fertility was also relatively low. Consequently, in thesecities, rural-to-urban migration was necessary for growth in order to compensate for the overall negative,or neutral, rate of natural increase. On the contrary, birth rates in African cities remain relatively high,while mortality has fallen to relatively low levels, due to colonial investments and the epidemiologicaltransition. African cities are “mushroom cities” whose growth is fueled by natural increase.6 If we knowthe urban growth rate and the urban rate of natural increase, we can estimate the respective contributionsof natural increase and what we call “residual migration”, the sum of the other components.

3.2 The "Killer Cities" of Industrial Europe

In this section we use historical data on the urban and rural demographic transitions in Industrial Europefrom 1700-1910 to explain the concept of “killer cities” (Williamson, 1990; Clark & Cummins, 2009).More precisely, for 7 representative European (or Neo-European) countries, we have been able to combineexisting data and collect historical data on the crude rates of birth and death for the largest city, theurban areas (considering all cities, including the largest city) and the rural areas for all available years.These countries are: England, France, Germany, the United States, Belgium, the Netherlands and Sweden.Where the data had not been collected yet by historians, we had to recreate it using old statistical reports,such as the annual Annuaires Statistiques de la France.7 While the data is imperfect, our study is to ourknowledge the first that uses consistent urban and rural series for 7 European countries over such a longperiod. We analyze England from 1700-1910 as a perfect example of these “killer cities”. Demographicpatterns in the English cities have been extensively described in the literature (Williamson, 1990; Clark &Cummins, 2009; Voigtländer & Voth, 2013b). This data spans two centuries.

Figure 7 plots the crude birth rate for London, urban areas, rural areas and England as a whole for allavailable years from 1700-1910. In 1700, the aggregate birth rate was 36 per 1,000 people. This birthrate was relatively low for England’s level of economic development. Most Western European countriesat that time were characterized by the “European Marriage Pattern”, in accordance with which women

6The term "mushroom" can be applied to growth or rapid expansion, seemingly overnight. Similarly, it is puzzling that Africancities grew so fast, almost “overnight”, and this fact is too often attributed to massive rural-to-urban migration. Conversely, wewant to show that natural increase has become the main source of urbanization in Africa.

7See Data Appendix for data sources.

9

married late and fertility was lower (Hajnal, 1965). What explains this specific pattern is unclear, butVoigtländer & Voth (2013a) shows how the Black Death in the 14th century had a long-term effect onmarriage and fertility patterns. In 1800, the first year for which we have separate data for urban and ruralareas, the urban birth rate (30.6) is 9.4 points lower than the rural birth rate (40.0). Birth rates did notdecrease before the 1880s. These patterns are confirmed by the analysis for the 6 other countries.

Similarly, Figure 8 plots the crude death rate for all available years from 1700-1910. In 1700, the aggregatedeath rate was 30 per 1,000 people, relatively low. Although England was poor, it was relatively less poorthan many European and non-European countries, and had better nutrition and health practices whichdecreased overall mortality. However, urban mortality was much higher, as exemplified by its evolution inLondon. In 1800, the urban death rate (32.5) was 11.6 points higher than the rural death rate (20.9). Highdensities inside walled (and non-walled) cities, industrial smoke, polluted water sources and unhygienicpractices all contributed to the urban penalty (Williamson, 1990; Voigtländer & Voth, 2009, 2013b). Eventhe urban wealthy faced high mortality risks (Clark & Cummins, 2009). The urban death rate remainedhigh until the 1860s. By 1900, the urban-rural mortality gap had disappeared and the aggregate deathrate had decreased to 15.2. These patterns are confirmed by the analysis for the six other countries.

In the 18th century, the aggregate natural increase rate was low or almost nil (see Figure 9). The naturalincrease rate was mostly negative for London, as a result of high mortality rates. Moreover, in the 19thcentury, it was high in the countryside, around 20 per 1,000 people, while it was very low in the cities,around 5 per 1,000 people. The urban rate only became positive after 1850. By 1900, the urban-rural gaphad almost disappeared. Appendix Figures A.1-A.6 show that these patterns are also present for the sixother countries. On average, the natural increase rate was 0.5% a year in the 19th century.

What are the respective contributions of natural increase and residual migration to urban growth in Eng-land from 1700-1910? Figure 10 shows the decomposition of urban growth into the two main components.Since the net international migration rates were almost zero, residual migration must have consisted ofrural-to-urban migration and urban reclassification. As expected, urban growth was driven by rural-to-urban migration, while the contribution of natural increase remained small until the late 19th century. Inother words, England could not have industrialized and urbanized without rural residents migrating tounhealthy urban environments. Results from other countries confirm these patterns. During the 1800-1910 period, the urban growth rate of Industrial Europe was 2% per year, while the urban rate of naturalincrease was 0.5%. The difference, 1.5%, was accounted for by residual migration. In Industrial Europe,net international migration was negative, so urban growth was driven by rural-to-urban migration (orreclassification). In the United States, international in-migration was high, so urban growth was driven byinternational-to-urban migration (and to a lesser extent by the other components).

3.3 The "Mushroom Cities" of Sub-Saharan Africa

In this section we use historical data on the urban and rural demographic transitions in Sub-SaharanAfrica from 1960-2010 to explain our concept our “mushroom cities”. More precisely, for 10 representativeAfrican countries, we have been able to collect panel data on the crude rates of birth and death for thelargest city, the urban areas (considering all cities, including the largest city) and the rural areas for eachdecade of the period. While data at the national level is readily available on the internet, separate datafor the urban and rural areas had to be recreated using census reports, census microdata, survey reports(such as the various Enquêtes Demographiques of the French Colonies of the late 1950s and early 1960s)and the reports of the more recent Demographic and Health Surveys that are available for many countries.8

While the data is imperfect, our study is to our knowledge the first that uses consistent urban and ruralseries for several African countries over such a long period.

Figure 11 plots the crude birth rate for the largest city, urban areas and rural areas of Africa from 1960-2010. At the time of independence, the aggregate crude birth rate was 49.5 per 1,000 people. Birth rateswere not significantly different between urban areas (49.0), the largest city (47.7) and rural areas (49.7).Birth rates of around 50 per 1,000 people are characteristic of pre-industrial societies and the first and

8See Data Appendix for data sources.

10

second stages of the demographic transition. The aggregate crude birth rate declined over time, but it isstill around 40 in the 2000s. However, birth rates declined relatively faster in the urban areas (36.2 in the2000s) and the largest city (32.3) than in the countryside (42.3). The birth rate is a function of the totalfertility rate and the number of women of reproductive age. However, at the cross-country level in Africa,the fertility rate is the main determinant of birth rates. For example, for 39 African countries for which wehave some data around 2000, the correlation coefficient between the two variables is 0.94 at the nationallevel, 0.85 for the urban areas and 0.92 for the rural areas (see Appendix Figure A.7). There are only fourcountries for which the urban birth rate is higher than the rural birth rate.9

Figure 12 plots the crude death rate from 1960-2010. At independence, the aggregate death rate was 25.0per 1,000 people. Death rates were significantly different across the largest city (15.5), urban areas (19.8)and rural areas (25.6). Death rates of around 25 are characteristic of the second stage of the demographictransition. The aggregate crude death rate declined over time, to around 13.8 in the 2000s. However,death rates declined relatively faster in the urban areas (36.2) and the largest city (32.3) than in thecountryside (42.3). The death rate is a function of the child mortality rate (0-5 years), the youth mortalityrate (5-15 years) and the adult mortality rate (15 and above years). However, at the cross-country levelin Africa, the child mortality rate is the main determinant of death rates.10 It has a 0.95 correlation withthe crude death rate at the national level for 47 African countries in 2000 (see Appendix Figure A.8).Unfortunately, we do not have separate data for urban and rural areas beyond the 10 countries of ourmain sample, but we have data on urban and rural child mortality for 39 Sub-Saharan African countriesfor the closest year to 2000. Urban child mortality is always lower, except in 4 countries.11

Historical evidence indicates that population growth was nil in Africa before 1900. Since the birth rate wasprobably around 50 per 1,000 people in 1900, the death rate must also have been around 50. In 1960, thebirth rate is still around 50, but the death rate is 25. Why was the death rate halved between 1900 and1960? Additional evidence indicates that mortality did not significantly decline before the 1930s. Indeed,during the 1900-1930 period, the population growth rate of Sub-Saharan Africa was 0.3% a year, vs. 2.2%a year in 1930-1960 and 2.7% a year in 1960-2010. For the growth rate to increase from 0.3% to to 2.2%a year in 1930-1960, given a stable birth rate of 50, the death rate must have decreased from 47 to 28between the two periods. What happened during the late colonial period that explains this sharp decreasein mortality? Although evidence is scarce, it seems that the colonizers invested significantly in health,educational and transportation infrastructure (e.g., missions, dispensaries, hospitals and railroads), whichled to better standards of living, as shown by anthropometric outcomes (Jedwab & Moradi, 2013). Forexample, Huillery (2009) shows that current health outcomes at the district level in former French WestAfrica were strongly determined by health investments in the colonial period. (Acemoglu & Johnson,2007) also show that the epidemiological transition of the mid 20th century and massive vaccinationcampaigns in the colonies resulted in widespread and significant declines in mortality. Colonial cities werecenters of diffusion of innovation, explaining why urban mortality was relatively lower.

However, the crude rates of natural increase evolved differentially across the largest city, urban areas andrural areas between 1960 and 2010. The aggregate crude rate of natural increase rate was 24.5 per 1,000people in the 1960s, and remained stabled over time (see Figure 13). All areas of Africa were alreadyexperiencing a demographic transition in the 1960s, due to lower death rates across all locations. But thegraph indicates that the demographic transition, as measured by the crude rate of natural increase, firsthappened in the largest city (with a rate of 32.2 in the 1960s), then in the other cities (31.8 in the 1970s)and then in the countryside (32.3 in the 1980s). In other words, cities boomed even before villages. Thedemographic transition has been “urban first”. These rates are now decreasing, as a result of the fertilitytransition, in line with the third stage of the demographic transition.

What are the respective contributions of natural increase and residual migration to urban growth in Sub-Saharan Africa from 1960-2010? Figure 14 shows the decomposition of urban growth into the two main

9However, for the four of them, the urban birth rate is less than 1 percentage point higher than the rural birth rate.10We focus on the period 1960-2010, during which HIV-related adult mortality was not a main concern. For instance, in South-

ern Africa where HIV prevalence rates are the highest in the world (Botswana, Lesotho, Namibia, South Africa and Swaziland),the average prevalence rate was 2.5% in 1990, 10.9% in 1995, 20.1% in 2000, 21.2% in 2005 and 20.7% 2010.

11However, excepting one country, the urban child mortality rate is not significantly lower than the rural child mortality rate.

11

components. Since the net international migration rates were very low, residual migration must haveconsisted of rural-to-urban migration and urban reclassification. As expected, the high urban growth rateof the 1960s was associated with a high residual migration rate (around 2.5% a year). Residual migrationkept decreasing in the following years, probably as a result of the political and economic crises of the1980s and 1990s, which drove down the urban growth rates. The contribution of natural increase wasalways larger than the contribution of residual migration, an upper bound to rural-to-urban migration.This confirms that the urban push was the main driver of urban growth in Africa, at least after the 1960s.Here, any urban pull or rural push increases the migration rate, and this first effect is then multiplied bynatural increase. The natural increase effect is very high, at 2.9% per year on average during the period,vs. 1.7% for residual migration. We find that a family of six migrants in 1960 would mechanically becomea family of twenty-five in 2010. We obtain similar patterns when using the largest city (see AppendixFigure A.9). This effect is even more visible in the 1980s when residual migration dropped.12

3.4 Comparison with other Developing Regions

In this section we use historical data on the urban demographic transition in other developing countries tocompare our results for Sub-Saharan Africa with the rest of the developing world. More precisely, for 22non-African developing countries, we have been able to collect panel data on the birth and death rates forthe urban and rural areas for each decade of the period.13 While the urban crude rate of natural increasewas 2.9% on average in Africa, it was 2.6% in the MENA region, 2.2% in the LAC region and 1.6% in Asia.When looking at subregions, the natural increase rate was high in Central America (2.5%), and less so inSouth America (2.0%) Southeast Asia (1.9%) and South Asia (1.8%). East Asia had the the lowest rate(1.1%), owing to the fertility transition in Japan and South Korea and the one-child policy in China.

As shown by Figure 4, the urban growth rate was also high in other developing regions in 1960-2010,but relatively less so in LAC and Asia. To what extent can these continental differences be explained bynatural increase rates? Appendix Figures 10, 11 and 12 display the decomposition of urban growth intonatural increase and residual migration for the three other regions. Clearly, urban growth in Asia wascloser to the European model, especially in the recent period, as urbanization was mostly associated withmigration (see Appendix Figure A.10). Conversely, in the MENA region, urban growth was closer to theAfrican model (see Appendix Figure A.11). This changed in the 2000s, when migration rates increasedas a result of economic growth and international migration from Asia (e.g., Filipino workers in the Gulfregion). Urban growth in the LAC region is also close to the African model, except migration rates werealready low in the 1960s (see Appendix Figure A.12).

As the natural increase rate decreases almost everywhere over time, urban growth is likely to deceleratein other developing regions, unless migration increases as in Asia. For example, in the MENA region, theurban crude rate of natural increase was 3.1% in the 1960s against 1.6% in the 2000s. In Central America,it was 3.4% in the 1960s vs. 1.7% in the 2000s. On the contrary, it remained high in Africa, at around2.6% in the 2000s. This indicates that African cities will keep growing due to natural increase for severaldecades, until African countries also complete their fertility transition. Urban fertility remains relativelyhigh in our African sample, while urban mortality is similar to the rest of the developing world.

4. IMPLICATIONS FOR URBAN POVERTY

In this section we develop a conceptual framework that allows us to explain the potential implications ofthe speed of urbanization and the different types of urban growth – migration vs. natural increase – onurban incomes. We then provide qualitative evidence of congestion effects in African cities and hypothesize

12These aggregated patterns hide significant differences across countries. Ivory Coast was one of the fastest-growing countriesbefore 1980, as it became the world leader of cocoa exports. Urban growth was 10% a year, while residual migration was 8%.Migration rates dropped as a result of the economic crisis in the 1980s. Urban growth remained high thereafter, because the rateof natural increase was 3% a year. In Ethiopia, the economic boom of the 1960s – with urban growth at 12% and migration at 9%– gave way to the Communist regime of Mengitsu. Migration dropped to 0% and urban growth was fueled by natural increase.

13See Data Appendix for a list of the countries that are included in the analysis and data sources.

12

that urban poverty increased and migration rates decreased as a result. Lastly, we discuss the role of publicpolicy in ensuring that urban growth and economic growth go hand in hand.

4.1 Conceptual Framework

4.1.1 Set-Up

We describe how the processes of economic development, internal migration, natural increase and urban-ization are interrelated. Suppose there are two economies. In the first economy (Economy A), urbanizationis associated with migration and rising urban incomes. In the second economy (Economy B), urbanizationis associated with natural increase and declining urban incomes, if urban labor supply increases fasterthan urban labor demand, as a result of urban congestion.

For the time being, suppose both economies have the same features. There is a rural labor market (R),where everyone receives the real rural subsistence wage WR/PR, with W being the nominal wage and Pbeing the level of prices.14 There is also an urban labor market (U), where the real urban wage WU/PUdepends on urban labor supply SU and demand DU . The production function in the urban sector is:

YU = AU KαU L1−αU (5)

where YU is the output of the urban sector, AU is an urban productivity parameter, KU is the stock of urbancapital and LU is the quantity of urban labor employed. KU is defined broadly as the stock of any type ofcapital: physical capital as in the Solovian model (machinery, buildings, computers, etc.), human capital,housing, roads, etc. Capital is inelastically supplied, so that 4KU < κ. In other words, capital cannot beaccumulated faster than κ. Urban wages are equal to the urban marginal product of labor:

WU

PU= αAU(

KU

LU)1−α (6)

The urban marginal product of labor is a positive function of the urban capital-labor ratio (KU/LU). Wealso assume that there are agglomeration economies that may increase urban productivity. AU is a functionof an exogenous productivity parameter ω and the size of the urban population LU :

AU = λLαU (7)

There are congestion effects if the population increases faster than the stock of capital 4LU >4KU , i.e. ifthe urban capital-labor ratio decreases. When population exogenously increases, more urban capital canbe accumulated, but not faster than κ. There is urban congestion if:

4LU > κ >4KU (8)

Thus, there are two contradictory effects of population growth on wages. Population growth increasesproductivity through agglomeration economies and decreases it through capital shallowing, i.e. urbancongestion. We do not explicitly model urban labor supply and demand. Urban labor supply depends onthe size of the urban population (which is inelastic, at least in the short term) and the urban-rural wagedifferential. As in the Harris-Todaro model, there is rural-to-urban migration (and an increase in urbanlabor supply, which is not inelastic) if the real urban wage is higher than the real rural wage:15

WR

PR<

WU

PU(9)

14We abstract from explicitly modeling the rural sector. As in the Lewisian model of urbanization, we assume that rural residentsmigrate to the cities as long as the urban wage is higher than the real rural subsistence wage. We hypothesize that adding anendogenous rural labor market would not change the main results of our theoretical exercise.

15Agents only compare rural and urban wages. However, in the Rosen-Roback model, agents compare the utility they get atdifferent locations, as cities could also have advantages or disadvantages in consumption. The “killer cities” of Industrial Europeoffered high wages to attract rural migrants to unhealthy urban environments. The “mushroom cities” of Africa were healthierthan the countryside, so the urban-rural utility gap could have been higher than the wage gap.

13

We assume that urban labor demand increases with urban labor productivity. If agglomeration effectsdominate the congestion effects, and labor productivity increases, the urban labor demand should increase.Depending on the response of urban labor supply, urban wages could also go up.

4.1.2 Economy A: Urbanization with Migration and Economic Development

Figure 15 provides a graphical representation of the urban labor market in Economy A. First, we assumethat the economy is poor and rural (LU is low) in period 0. There is no urban-rural wage differential, sothe urban wage equals the rural subsistence wage. What happens if there is a one-time exogenous increasein productivity, such as technological change during the Industrial Revolution (4λ > 0)?

Urban labor demand curve shifts to the right and the urban wage increases. The size of the urban popu-lation is inelastic in the short term, but urban labor supply is not, thanks to migration. Thus, if the wagegoes up, the number of migrants increases, and urban labor supply shifts to the right. If labor supply andmigration do not entirely adjust to the shift in labor demand, the urban wage is higher than the rural wageat the end of period 1. The economy is now wealthier and more urban.16

Now assume that the increase in urban growth in period 1 raises urban productivity in period 2, becauseagglomeration effects dominate congestion effects, as shown in Figure 16. Labor demand curve shirtsfurther to the right and the urban wage goes up. This attracts even more migrants, so the labor supplycurve also shifts further to the right. If labor supply does not rise as fast as labor demand, the urban wagerises further. In this economy, a one-time increase in productivity leads to the following sequence of events:increase in the urban wage, migration, expansion of the urban population, rise in productivity, increase inthe urban wage, etc. Urbanization is associated with migration and economic development.

4.1.3 Economy B: Urbanization without Migration and Economic Development

Figure 17 provides a graphical representation of the urban labor market in Economy B. First, we assumethat Economy B is not different from Economy A in periods 0 and 1. A one-time exogenous productivityshock raises the urban wage and encourages rural-to-urban migration (see Figure 15). However, we nowposit that the increase in the urban population in period 1 also leads to an exogenous (and unexpected)increase in the urban labor supply in period 2, as a result of natural increase (see Figure 17).

As shown by equation (6), this exogenous population growth increases productivity through agglomer-ation economies and decreases it through congestion effects (the urban capital-labor ratio), if capital isinelastically supplied. In Economy A, we assume agglomeration effects strongly dominate congestioneffects. However, in Economy B, we assume that the growth of the labor supply exceeds the maximalchange in the stock of capital, in line with equation (8). Congestion effects kick in. As a result, even iflabor productivity increases on net, and the labor demand curve shifts to the right, the increase in theurban labor supply may be too large to be “absorbed” without reducing the urban wage. In this economy,a one-time increase in productivity leads to the following sequence of events: rise in the urban wage,migration, expansion of the urban population, exogenous natural increase, additional expansion of theurban population, smaller increase in productivity, decline in the urban wage, etc.

4.1.4 Comparison of the Economies A and B

This conceptual framework allows us to obtain a set of predictions about the responses of the urban wage,migration and urbanization to an one-time exogenous productivity shock in both economies. Figure 18shows the evolution of these three variables over time in Economy A and Economy B.

16Although we do not explicitly model the migration decision, we assume that there are some constraints on migration whichensure that urban labor supply does not adjust instantaneously. First, there could be migration costs in the short term. Second, ifthere are heterogenous agents and people have different “intrinsic” preferences for rural (or urban) living, the urban wage mayhave to increase a lot (or a little) to fuel migration, as agents compare marginal utilities and not just wages.

14

In the former, the urban wage continuously increases, as agglomeration effects always dominate the con-gestion effects. The urban-rural wage differential leads to increasing migration and urbanization overtime. In the latter, the urban wage increases in the short run, because the migration effect is multiplied bythe natural increase in the medium run. Urban population growth is excessive and urban incomes decline,so that migration rates slow down, although the economy keeps urbanizing. In the long run, urbanizationis associated with declining urban incomes. This result comes from our assumption that urban capital isinelastically supplied (4KU < κ), so that congestion effects kick in if the speed of urbanization is too fast.In the next section, we discuss why we think this assumption is credible.

4.2 Qualitative Evidence on Congestion Effects [TO BE COMPLETED]

The main hypothesis of our conceptual framework is that urban real wages have been relatively decliningin African countries over time, especially compared to the 1950s and 1960s, when economic growth washigh. Urban wages may have decreased because of declining nominal wages (WU) and increasing urbanprices (PU). If the urban-rural wage (or marginal utility) differential has decreased over time, we shouldalso observe a reduction in rural-to-urban migration rates.

• Income and migration: Due to the paucity of data for African countries, we cannot test whetherthe urban-rural income gap has been declining over time. For the recent period, Gollin, Lagakos &Waugh (2011) find that value added per worker was on average twice as high in the non-agriculturalsector as in the agricultural sector. If the former is mostly urban and the latter is mostly rural, theurban-rural income gap is still greater than 1 in developing countries, particularly in Ghana and IvoryCoast for which they use household survey data. Yet Young (2013) does not find an urban-rural wagegap once he controls for the (observable and unobservable) skill intensity of production in urbanand rural areas. Similarly, Bocquier, Madise & Zulu (2011) do not find an urban advantage in childsurvival once they control for known demographic and socioeconomic correlated variables. There issome evidence that the urban-rural gap was significant before. For example, Jamal & Weeks (1988,1993) and Potts (1995) show how the debt crisis of the 1970s-1980s and the structural adjustmentprograms of the 1980s-1990s reduced the rural-urban income gap, which was high at the time ofindependence.17 Therefore, the debt crisis, and not urban congestion, was the main determinantof declining urban incomes. However, urban congestion may explain why migration rates remainedlow in the 2000s, as shown by Figure 14. Residual migration is still calculated to be above 1% peryear, but this is an upper bound to rural migration as it also includes reclassification. Additionally,there are now many examples of circular migration, wherein people migrate from the cities to thevillages or from large cities to smaller towns (Beauchemin & Bocquier, 2004; Potts, 2009). We planto use census microdata from IPUMS and survey data from the Demographic and Health Surveys(DHS) to investigate whether rural-to-urban migration rates have decreased over time, and whetherthese were nil in the 2000s. As the DHS often reports the birth place of individuals (urban vs. rural),we will be able to study migration rates for different cohorts.

• Housing prices and quality: Any comparison of monetary outcomes should take into accountprices, as prices are likely to have risen in urban areas, especially housing prices. As the popula-tion of African cities doubled every 15 years, the housing stock had to increase rapidly to match thedemand for housing. If housing supply did not increase as fast as housing demand, the number ofpeople per room may have gone up or slums may have worsened. According to our theory, slums donot grow because they receive a massive influx of poor rural migrants that flock to cities. They growbecause of natural increase and the fact that the housing supply is inelastic, at least in the short run.In other words, the slums “mushroom”. According to UN-Habitat (2003), 72% of urban residentswere living in slums in Africa in 2001, and things could get worse. For example, the report indicatesthat (p.25): “Most cities in sub-Saharan Africa [...] showed considerable housing stress, with rents

17For example, Potts (1995) writes (p.1): “Many public sector workers [...] were reduced to incomes which only covered afraction of the absolute necessities of existence, or, worse still, lost their jobs. Urban service provision has also declined, oftendramatically. The gap between real rural incomes and real urban incomes has often narrowed considerably. It appears thatthe rate of urban growth in some African countries has slowed considerably, and there is also some evidence that new forms of‘reverse migration’ from urban to rural areas have occurred.”

15

and prices rising substantially while incomes fell, probably corresponding to higher occupancy rates.In addition, slum areas increased in most cities, and the rate of slum improvement was very slow ornegligible in most places.” We plan to use data from the United Nations and the DHS to see if theshare of urban residents living in slums has been increasing over time (from the mid-1980s to thelate 2000s, as there are no data for the earlier period).18 We also want to test if the evolution of the“composition” of slum areas (rural migrants vs. urban natives) validates our theory. We predict thatmost residents of slums were actually born in these cities.19

• Informal employment and unemployment: While urban economies were relatively formal at inde-pendence, most urban residents now work in the informal sector (ILO, 2009). Youth unemploymentrates are also relatively high in Africa. While the debt crisis clearly accelerated the informalization ofAfrican economies, natural increase was also a contributing factor. Urban labor markets were prob-ably not able to absorb these massive (and unexpected) labor supply shocks. In a highly regulatedenvironment, the informal sector acted as a residual sector for this excess labor. We plan to use somecensus data to document the informalization of African urban economies.

• Human capital deficiency: As the population of African cities doubled every 15 years, the numberof educational and health facilities had to increase rapidly to match the demand for human capital.However, the education and health sectors are often highly regulated in the cities. Governmentshave been unable to keep up with the population growth. They need to invest in new facilities andtrain and hire new specialized workers (e.g., teachers, physicians, etc.). Though there are signs ofthis inadequate infrastructure, such as overcrowded schools and hospitals, there is little empiricalevidence on this type of congestion in African cities.

• Road congestion: Traffic jams have become a major issue in Africa. The outward spreading ofAfrican cities, the lack of efficient public transport and an increase in car ownership rates all con-tribute to rising road congestion (UN-Habitat, 2008). Although there is anecdotal evidence thatroad congestion may have increased over time, unfortunately we will not be able to show empiricalevidence on this type of urban congestion.

4.3 Policy Implications [TO BE COMPLETED]

If African cities grew too fast – at 4.7% per year – and urban welfare decreased as a result of urban con-gestion, what can be done about it? In our conceptual framework, there are three possibilities if the speedis very fast: high economic growth rates, a reduction in urban fertility or better urban planning.

To begin, a rapid speed of urbanization is only a problem if urban incomes are not increasing independentlyof the urbanization process. In the 19th century, New York was growing at 4.3% per year while theother U.S. cities had annual growth rates of around 3% (growth rates were below 3% in Europe). Thesecities were also affected by congestion. However, economic growth was high, as a result of technologicalprogress (λ in our conceptual framework), and that is why these cities attracted migrants in the first placeand grew. Urban growth was the result of higher urban incomes, and congestion effects were not largeenough to offset the initial gains. However, African countries urbanized as a result of natural increase, andtechnological progress is a relatively less important factor in their cities today. So African countries mustcope with the rapid growth of their cities, without capturing the benefit of technological growth.

Then, any population growth slowdown would contribute to increasing the capital-labor ratio, if conges-tion effects do not kick in (4LU < κ). This could be achieved through a reduction in the urban fertilityrate. Urban fertility is still very high in Africa. For example, the average urban birth rate was 36.2 per

18A related question is whether the slums of Africa today are worse than the slums of Industrial Europe. Slums consisted ofsmall brick houses in England and tenements in the United States. On the contrary, slums in Africa consist of mostly shacks. Thisvariation in the type of slums could be due to differing weather conditions between developed countries and Africa. However,slums are mostly shack settlements in South Africa, where the climate is more continental. Answering this question will bedifficult, but we plan to collect historical data on the extent and the type of slums across space and time.

19A related question is whether natural increase happens disproportionately in slum areas or non-slum areas of African cities.For example, Fink, Günther & Hill (2013) show that mortality is higher in non-slum areas than in slum areas. However, fertilitycould also be higher, and the net effect on natural increase is unknown.

16

1,000 people for the ten countries of our sample in the 2000s. This is still higher than the birth ratesin Industrial Europe during the 19th century (around 30) and other developing countries today (around10-20). Given an urban death rate of 11.3, reducing the urban birth rate from 36.2 to 20 would lead to anatural increase rate of 8.7. With 1.7% migration, this would result in an urbanization rate of 2.6%, whichis close to what we had in Europe before and what we have in Asia now. Our analysis emphasizes the roleof urban family planning policies, as rapid demographic growth is also an urban problem.20

Lastly, even without economic growth and without a reduction in fertility, better urban planning could per-mit an internalization of (negative) urban externalities. There are several possible approaches. First, theremodeling of Paris by Baron Haussmann in the 1850s is a perfect example of the authoritarian approach.He cleared the narrow medieval streets of the capital in favor of broad boulevards. This transformationincreased the standard of living of the Parisians in the later period. Though this approach was undoubt-edly beneficial in the long-run, it is highly controversial as a policy model due to its high short-run societalcosts. Second, many cities were planned as a result of (unplanned) creative destruction. For example, afew European cities and many American cities were rebuilt in a better way after a Great Fire (e.g. New Yorkin 1776, Chicago 1871, Boston 1872, San Francisco 1906 and London in 1940). City fires in developingcountries today are much less destructive, for various reasons. Houses are built with cement and shacksare built with metal sheets, rather than wood. Fire departments are also more efficient. Third, local urbanrenewal projects are examples of a more decentralized approach, whether they are implemented by localgovernments or private promoters. Collins & Shester (2013) use an instrumental variable strategy to showthat urban renewal projects had positive effects on income, property values and population in the U.S.However, in developing countries, the absence of strong private markets and corruption may decrease theeconomic returns to urban renewal projects and resettlement programs. Thus, without any improvementin urban (not just national) institutions, urban congestion is likely to remain a major issue.

5. CONCLUSION

This paper documents several new facts regarding the processes of economic development, internal migra-tion and urbanization. In particular, countries that urbanized through natural increase and whose speedof urbanization was very fast may have seen an increase in urban poverty, as a result of urban congestion.We develop a conceptual framework that is able to capture both the typical pattern of urbanization, mi-gration and economic development seen across much of Industrial Europe and developing Asia as well asurbanization without migration and development as seen in many Sub-Saharan African countries.

This work suggests that urbanization per se is not necessarily conducive to economic growth. One strandof the literature posits that agglomeration promotes economic growth. Another strand of the literatureshows that countries can urbanize without growth. By also urbanizing through natural increase, and notjust through migration, African countries are disproportionately affected by urban congestion, which leadsto declining urban incomes over time. In other words, urbanization with growth encourages migration,and this first effect is then multiplied by natural increase, which may produce urbanization without growth.The source and the speed of urbanization appear to be important factors in future development.

This paper leaves several open questions. The first is why African countries and African cities did notcomplete their fertility transition earlier. Indeed, urban fertility remains high, and African cities are likelyto grow further in the coming decades. A second question that we leave unanswered is why the speed ofurbanization was relatively lower in some African countries, e.g. South Africa, and whether their cities arenow better able to benefit from agglomeration economies and solve urban congestion.

20Urban mortality rates have recently increased as a result of HIV in a few African countries. In Southern Africa where HIVprevalence rates are the highest in the world (Botswana, Lesotho, Namibia, South Africa and Swaziland), the aggregate deathrate increased from 9 per 1,000 people in 1990 to 15 in 2010. The aggregate birth rate decreased from 28 to 23 during the sameperiod. Thus, the aggregate rate of natural increase decreased by 11. We do not have separate data for urban and rural areas, butnatural increase probably decreased by more than 11 (1.1%) in the cities. This could also explain why urban growth rates wererelatively lower on average in these countries than in the rest of Africa. As in Young (2005), it is possible that the HIV epidemicshad a relatively positive effect on urban per capita incomes, the channel here being a lower speed of urbanization.

17

REFERENCES

Acemoglu, Daron, and Simon Johnson. 2007. “Disease and Development: The Effect of Life Expectancyon Economic Growth.” Journal of Political Economy, 115(6): 925–985.

Acemoglu, Daron, Simon Johnson, and James A. Robinson. 2005. “The Rise of Europe: Atlantic Trade,Institutional Change, and Economic Growth.” American Economic Review, 95(3): 546–579.

Ades, Alberto F, and Edward L Glaeser. 1995. “Trade and Circuses: Explaining Urban Giants.” The Quar-terly Journal of Economics, 110(1): 195–227.

Alvarez-Cuadrado, Francisco, and Markus Poschke. 2011. “Structural Change Out of Agriculture: LaborPush versus Labor Pull.” American Economic Journal: Macroeconomics, 3(3): 127–58.

Ashraf, Quamrul, and Oded Galor. 2011. “Dynamics and Stagnation in the Malthusian Epoch.” AmericanEconomic Review, 101(5): 2003–41.

Ashraf, Quamrul H., David N. Weil, and Joshua Wilde. 2011. “The Effect of Interventions to ReduceFertility on Economic Growth.” National Bureau of Economic Research Working Paper 17377.

Bairoch, Paul. 1988. Cities and Economic Development: Frow the Dawn of History to the Present. Chicago:The University of Chicago Press.

Barrios, Salvador, Luisito Bertinelli, and Eric Strobl. 2006. “Climatic Change and Rural-Urban Migra-tion: The Case of Sub-Saharan Africa.” Journal of Urban Economics, 60(3): 357–371.

Bates, Robert. 1981. Markets and States in Tropical Africa: The Political Basis of Agricultural Policies.Berkeley: University of California Press.

Beauchemin, Cris, and Philippe Bocquier. 2004. “Migration and Urbanisation in Francophone WestAfrica: An Overview of the Recent Empirical Evidence.” Urban Studies, 41(11): pp. 2245–2272.

Becker, Gary S., Edward L. Glaeser, and Kevin M. Murphy. 1999. “Population and Economic Growth.”American Economic Review, 89(2): 145–149.

Bleakley, Hoyt. 2007. “Disease and Development: Evidence from Hookworm Eradication in the AmericanSouth.” The Quarterly Journal of Economics, 122(1): 73–117.

Bleakley, Hoyt. 2010. “Malaria Eradication in the Americas: A Retrospective Analysis of Childhood Expo-sure.” American Economic Journal: Applied Economics, 2(2): 1–45.

Bleakley, Hoyt, and Fabian Lange. 2009. “Chronic Disease Burden and the Interaction of Education,Fertility, and Growth.” The Review of Economics and Statistics, 91(1): 52–65.

Bocquier, Philippe, Nyovani Madise, and Eliya Zulu. 2011. “Is There an Urban Advantage in ChildSurvival in Sub-Saharan Africa? Evidence From 18 Countries in the 1990s.” Demography, 48(2): 531–558.

Caselli, Francesco, and Wilbur John Coleman II. 2001. “The U.S. Structural Transformation and Re-gional Convergence: A Reinterpretation.” Journal of Political Economy, 109(3): 584–616.

Clark, Gregory, and Neil Cummins. 2009. “Urbanization, Mortality, and Fertility in Malthusian England.”American Economic Review, 99(2): 242–47.

Collins, William J., and Katharine L. Shester. 2013. “Slum Clearance and Urban Renewal in the UnitedStates.” American Economic Journal: Applied Economics, 5(1): 239–73.

Cutler, David, Winnie Fung, Michael Kremer, Monica Singhal, and Tom Vogl. 2010. “Early-Life MalariaExposure and Adult Outcomes: Evidence from Malaria Eradication in India.” American Economic Jour-nal: Applied Economics, 2(2): 72–94.

Davis, James C., and J. Vernon Henderson. 2003. “Evidence on the political economy of the urbanizationprocess.” Journal of Urban Economics, 53(1): 98–125.

18

Diamond, Jared. 1997. Guns, Germs, and Steel: The Fates of Human Societies. New York: W. W. Norton &Company.

Duranton, Gilles. 2008. “Viewpoint: From Cities to Productivity and Growth in Developing Countries.”Canadian Journal of Economics, 41(3): 689–736.

Fay, Marianne, and Charlotte Opal. 2000. “Urbanization Without Growth : a Not-so-Uncommon Phe-nomenon.” The World Bank Policy Research Working Paper Series 2412.

Felkner, John S., and Robert M. Townsend. 2011. “The Geographic Concentration of Enterprise in De-veloping Countries.” The Quarterly Journal of Economics, 126(4): 2005–2061.

Fink, Günther, Isabel Günther, and Kenneth Hill. 2013. “Urban Mortality Transitions: The Role ofSlums.” Program on the Global Demography of Aging PGDA Working Papers 9913.

Geopolis. 2012. Africapolis I: Western Africa, and Africapolis II : Central and Eastern Africa. Paris: SEDET,CNRS-Université Paris-Diderot.

Glaeser, Edward L., and Joshua D. Gottlieb. 2009. “The Wealth of Cities: Agglomeration Economies andSpatial Equilibrium in the United States.” Journal of Economic Literature, 47(4): 983–1028.

Gollin, Douglas, David Lagakos, and Michael E. Waugh. 2011. “The Agricultural Productivity Gap inDeveloping Countries.”

Gollin, Douglas, Remi Jedwab, and Dietrich Vollrath. 2013. “Urbanization with and without Industrial-ization.” Unpublished manuscript, Oxford University Department of International Development.

Gollin, Douglas, Stephen L. Parente, and Richard Rogerson. 2002. “The Role of Agriculture in Devel-opment.” American Economic Review, 92(2): 160–164.

Greenstone, Michael, Richard Hornbeck, and Enrico Moretti. 2010. “Identifying AgglomerationSpillovers: Evidence from Winners and Losers of Large Plant Openings.” Journal of Political Economy,118(3): 536–598.

Hajnal, John. 1965. “European Marriage Pattern in Historical Perspective.” In Population in History. , ed.D.V. Glass and D.E.C. Eversley. London: Arnold.

Hansen, Gary D., and Edward C. Prescott. 2002. “Malthus to Solow.” American Economic Review,92(4): 1205–1217.

Harris, John R., and Michael P. Todaro. 1970. “Migration, Unemployment & Development: A Two-SectorAnalysis.” American Economic Review, 60(1): 126–42.

Henderson, J. Vernon. 2005. “Urbanization and Growth.” In Handbook of Economic Growth. Vol. 1, , ed.P. Aghion and S. Durlauf, Chapter 24, 1543–1591. Elsevier.

Henderson, J. Vernon. 2010. “Cities and Development.” Journal of Regional Science, 50(1): 515–540.

Huillery, Elise. 2009. “History Matters: The Long-Term Impact of Colonial Public Investments in FrenchWest Africa.” American Economic Journal: Applied Economics, 1(2): 176–215.

ILO. 2009. The Informal Economy in Africa: Promoting Transition to Formality: Challenges and Strategies.Geneva: International Labour Organization.

Jamal, Vali, and John Weeks. 1988. “The Vanishing Rural-Urban Gap in Sub-Saharan Africa.” Interna-tional Labour Review, 127(3): pp. 271–292.

Jamal, Vali, and John Weeks. 1993. Africa Misunderstood, or Whatever Happened to the Rural-Urban Gap?London: MacMillan.

Jedwab, Remi. 2013. “Urbanization without Industrialization: Evidence from Consumption Cities inAfrica.” Unpublished manuscript, Department of Economics, George Washington University.

19

Jedwab, Remi, and Alexander Moradi. 2013. “Colonial Investments and Long-Term Development: Evi-dence from African Railroads.” Unpublished manuscript, Department of Economics, George WashingtonUniversity.

Kremer, Michael. 1993. “Population Growth and Technological Change: One Million B.C. to 1990.” TheQuarterly Journal of Economics, 108(3): pp. 681–716.

Lagerlöf, Nils-Petter. 2003. “Mortality and Early Growth in England, France and Sweden.” ScandinavianJournal of Economics, 105(3): 419–440.

Lagerlöf, Nils-Petter. 2006. “The Galor-Weil Model Revisited: A Quantitative Exercise.” Review of EconomicDynamics, 9(1): 116–142.

Lewis, Arthur. 1954. “Economic Development with Unlimited Supplies of Labour.” The Manchester School,22(2): 139–191.

Lipton, Michael. 1977. Why Poor People Stay Poor: Urban Bias in World Development. Cambridge: HarvardUniversity Press.

Lucas, Robert E. 2004. “Life Earnings and Rural-Urban Migration.” Journal of Political Economy,112(S1): S29–S59.

Matsuyama, Kiminori. 1992. “Agricultural Productivity, Comparative Advantage, and Economic Growth.”Journal of Economic Theory, 58(2): 317–334.

McKinsey. 2011. Lions on the Move: The Progress and Potential of African Economies. London : Mc KinseyGlobal Institute.

Michaels, Guy, Ferdinand Rauch, and Stephen J. Redding. 2012. “Urbanization and Structural Trans-formation.” The Quarterly Journal of Economics, 127(2): 535–586.

Nunn, Nathan, and Nancy Qian. 2011. “The Potato’s Contribution to Population and Urbanization: Evi-dence from an Historical Experiment.” The Quaterly Journal of Economics, 126(2).

Overman, Henry, and Anthony J. Venables. 2005. “Cities in the Developing World.” Centre for EconomicPerformance, LSE CEP Discussion Papers dp0695.

Poelhekke, Steven. 2010. “Urban Growth and Uninsured Rural Risk: Booming Towns in Bust Times.”Journal of Development Economics, In Press.

Potts, Deborah. 1995. “Shall We Go Home? Increasing Urban Poverty in African Cities and MigrationProcesses.” The Geographical Journal, 161(3): pp. 245–264.

Potts, Deborah. 2009. “The Slowing of sub-Saharan Africa’s Urbanization: Evidence and Implications forUrban Livelihoods.” Enviroment and Urbanization, 21(1): pp. 253–258.

Potts, Deborah. 2012. “Challenging the Myths of Urban Dynamics in Sub-Saharan Africa: The Evidencefrom Nigeria.” World Development, 40(7): 1382–1393.

Ravallion, Martin, Shaohua Chen, and Prem Sangraula. 2007. “New Evidence on the Urbanization ofGlobal Poverty.” Population and Development Review, 33(4): 667–701.

Rosenthal, Stuart S., and William C. Strange. 2004. “Evidence on The Nature and Sources of Agglomer-ation Economies.” In Handbook of Regional and Urban Economics. Vol. 4, , ed. J. V. Henderson and J. F.Thisse, Chapter 49, 2119–2171. Elsevier.

Satterthwaite, David. 2007. “The Transition to a Predominantly Urban World and Its Underpinnings.”Human Settlements Working Paper Series Urban Change, 4.

Schultz, Theodore W. 1953. The Economic Organization of Agriculture. New York: McGraw-Hill.

UN-Habitat. 2003. The Challenge of Slums - Global Report on Human Settlements. London and Sterling VA:Earthscan Publication Ltd.

20

UN-Habitat. 2008. The State of the African Cities. Nairobi: United Nations Human Settlements Programme.

United Nations. 2011. World Urbanization Prospects: The 2011 Revision. New York: United Nations, De-partment of Economic and Social Affairs.

Venables, Anthony J. 2010. “Economic Geography and African Development.” Papers in Regional Science,89(3): 469–483.

Voigtländer, Nico, and Hans-Joachim Voth. 2006. “Why England? Demographic factors, structuralchange and physical capital accumulation during the Industrial Revolution.” Journal of Economic Growth,11(4): 319–361.

Voigtländer, Nico, and Hans-Joachim Voth. 2009. “Malthusian Dynamism and the Rise of Europe: MakeWar, Not Love.” American Economic Review, 99(2): 248–54.

Voigtländer, Nico, and Hans-Joachim Voth. 2013a. “How the West ’Invented’ Fertility Restriction.” Forth-coming in the American Economic Review.

Voigtländer, Nico, and Hans-Joachim Voth. 2013b. “The Three Horsemen of Riches: Plague, War andUrbanization in Early Modern Europe.” Forthcoming in the Review of Economic Studies.

Vollrath, Dietrich. 2011. “The agricultural basis of comparative development.” Journal of EconomicGrowth, 16(4): 343–370.

Weil, David N. 2007. “Accounting for The Effect of Health on Economic Growth.” The Quarterly Journal ofEconomics, 122(3): 1265–1306.

Weil, David N., and Oded Galor. 1999. “From Malthusian Stagnation to Modern Growth.” AmericanEconomic Review, 89(2): 150–154.

Weil, David N., and Oded Galor. 2000. “Population, Technology, and Growth: From Malthusian Stagna-tion to the Demographic Transition and Beyond.” American Economic Review, 90(4): 806–828.

Williamson, Jeffrey. 1990. Coping with City Growth During the British Industrial Revolution. Cambridge:Cambridge University Press.

World Bank. 2009. World Development Report 2009: Reshaping Economic Geography. Washington DC:World Bank Publications.

Young, Alwyn. 2005. “The Gift of the Dying: The Tragedy of Aids and the Welfare of Future AfricanGenerations.” The Quarterly Journal of Economics, 120(2): 423–466.

Young, Alwyn. 2013. “Inequality, the Urban-Rural Gap and Migration.” Unpublished manuscript, LondonSchool of Economics, Department of Economics.

21

Figure 1: Urbanization Rates for Industrial Europe, 1700-1950

Notes: This figure plots the average urbanization rate (%) for “Industrial Europe” in 1700-1950. Industrial Europe includesthe following Western European countries or regions, since we only have consistent historical data for 12 spatial units: UnitedKingdom (incl. England, Wales, Northern Ireland and Scotland), Ireland, France, Belgium, The Netherlands, Germany, Austria(incl. Hungary), Switzerland, Italy, Spain, Portugal and Scandinavia. Industrial Europe also includes the United States in ouranalysis, as one example of a Neo-European country. Averages are estimated using the population weights for the same year. SeeData Appendix for data sources.

Figure 2: Urbanization Rates for Four Developing Regions, 1950-2010

Notes: This figure plots the average urbanization rate (%) for four groups of countries in 1950-2010: Africa (46 countries), Asia(30 countries), Latin America and the Caribbean LAC (26 countries) and Middle-East and North Africa MENA (17 countries).Averages are estimated using the population weights for the same year. See Data Appendix for data sources.

22

Figure 3: Urban Growth Rates for Industrial Europe, 1700-1950

Notes: This figure plots the average annual urban growth rate (%) for “Industrial Europe” in 1700-1950. Industrial Europeincludes the following Western European countries or regions, since we only have consistent historical data for 12 spatial units:United Kingdom (incl. England, Wales, Northern Ireland and Scotland), Ireland, France, Belgium, The Netherlands, Germany,Austria (incl. Hungary), Switzerland, Italy, Spain, Portugal and Scandinavia. Industrial Europe also includes the United States inour analysis, as one example of a Neo-European country. Averages are estimated using the population weights for the same year.See Data Appendix for data sources.

Figure 4: Urban Growth Rates for Four Developing Regions, 1950-2010

Notes: This figure plots the average annual urban growth rate (%) for selected periods for four groups of countries in 1950-2010:Africa (46 countries), Asia (30 countries), Latin America and the Caribbean LAC (26 countries) and Middle-East and North AfricaMENA (17 countries). Averages are estimated using the population weights for the same year. See Data Appendix for datasources.

23

Figure 5: Urban Growth Rates for All African Countries, 1950-2010

Notes: This map displays the annual urban growth rate (%) for 46 African countries in 1950-2010. See Data Appendix for datasources.

Figure 6: Growth Rate of the Largest City for Developing Regions, 1950-2010

Notes: This figure plots the annual growth rate of the largest city (%) for selected periods for four groups of countries in 1950-2010: Africa (46 countries), Asia (30 countries), Latin America and the Caribbean LAC (26 countries) and Middle-East and NorthAfrica MENA (17 countries). The largest city is not necessarily the capital city (e.g., Abidjan instead of Yamoussoukro in IvoryCoast). Averages are estimated using the population weights for the same year. See Data Appendix for data sources.

24

Figure 7: Crude Birth Rates for England, Urban vs. Rural, 1700-1910

Notes: This figure plots the crude birth rate (per 1,000 people) for England, rural England, urban England (considering all cities,including London) and London in 1700-1910. The data is missing for London in 1700-1720. The data is missing for the ruraland urban areas of England in 1700-1790. See Data Appendix for data sources.

Figure 8: Crude Death Rates for England, Urban vs. Rural, 1700-1910

Notes: This figure plots the crude death rate (per 1,000 people) for England, rural England, urban England (considering all cities,including London) and London in 1700-1910. The data is missing for London in 1700-1720. The data is missing for the ruraland urban areas of England in 1700-1790. See Data Appendix for data sources.

25

Figure 9: Crude Rates of Natural Increase for England, 1700-1910

Notes: This figure plots the crude rate of natural increase (per 1,000 people) for England, rural England, urban England (con-sidering all cities, including London) and London in 1700-1910. The crude rate of natural increase is the difference between thecrude birth rate and the crude death rate. The data is missing for London in 1700-1720. The data is missing for all urban andrural areas in 1700-1790. The average urban crude rate of natural increase is 0.5% in 1800-1910 (for available years only). SeeData Appendix for data sources.

Figure 10: Migration, Natural Increase and Urban Growth, England 1700-1910

Notes: This figure plots the decomposition of annual urban growth (%) into annual natural increase (%) and annual “residualmigration” (%). Residual migration includes rural-to-urban migration, international migration and urban reclassification. If thereis some urban reclassification, residual migration is an upper bound to migration. Since we do not have data on the urban deathrate in 1700-1790, we use the death rate of London instead. See Data Appendix for data sources.

26

Figure 11: Crude Birth Rates for Africa, Urban vs. Rural, 1960-2010

Notes: This figure plots the crude birth rate (per 1,000 people) for Africa, rural Africa, urban Africa (considering all cities,including the largest city of each country) and the largest city of each country in 1960-2010. We use historical census andsurvey data that we have collected for 10 representative Sub-Saharan African countries: Burkina-Faso, Central African Republic,Ethiopia, Ghana, Ivory Coast, Kenya, Madagascar, Malawi, Mali and Senegal. See Data Appendix for data sources.

Figure 12: Crude Death Rates for Africa, Urban vs. Rural, 1960-2010

Notes: This figure plots the crude death rate (per 1,000 people) for Africa, rural Africa, urban Africa (considering all cities,including the largest city of each country) and the largest city of each country in 1960-2010. We use historical census andsurvey data that we have collected for 10 representative Sub-Saharan African countries: Burkina-Faso, Central African Republic,Ethiopia, Ghana, Ivory Coast, Kenya, Madagascar, Malawi, Mali and Senegal. See Data Appendix for data sources.

27

Figure 13: Crude Rates of Natural Increase for Africa, 1960-2010

Notes: This figure plots the crude rate of natural increase (per 1,000 people) for Africa, rural Africa, urban Africa (considering allcities, including the largest city of each country) and the largest city of each country in 1960-2010. We use historical census andsurvey data that we have collected for 10 representative Sub-Saharan African countries: Burkina-Faso, Central African Republic,Ethiopia, Ghana, Ivory Coast, Kenya, Madagascar, Malawi, Mali and Senegal. The crude rate of natural increase is equal to thecrude birth rate minus the crude death rate. The average urban crude rate of natural increase is 2.9% in 1960-2010. See DataAppendix for data sources.

Figure 14: Migration, Natural Increase and Urban Growth, Africa, 1960-2010

Notes: This figure plots the decomposition of annual urban growth (%) into annual natural increase (%) and annual residualmigration (%). Residual migration includes rural-to-urban migration, international migration and urban reclassification. If thereis some urban reclassification, residual migration is an upper bound to migration. We use historical census and survey data thatwe have collected for 10 representative Sub-Saharan African countries: Burkina-Faso, Central African Republic, Ethiopia, Ghana,Ivory Coast, Kenya, Madagascar, Malawi, Mali and Senegal. See Data Appendix for data sources.

28

Figure 15: Urbanization with Migration and Economic Development, Period 1

LU

WU

DU,0

SU,0

WR

PU

PR

DU,1

SU,1

Notes: This figure plots the effect of a productivity shock (λ) on the urban labor demand (DU ,1). The urban-rural wage differentialis higher than 1, which increases migration. If urban labor supply does not increase as fast as urban labor demand, urban wagesreach a higher level.

Figure 16: Urbanization with Migration and Economic Development, Period 2

LU

WU

DU,0

SU,0

WR

PU

PR

DU,1

SU,1

DU,2

SU,2

Notes: Agglomeration effects (LγU ) further increase urban labor demand (DU ,2). If urban labor supply does not increase as fast asurban labor demand, urban wages reach an even higher level.

29

Figure 17: Urbanization without Migration and Economic Development

LU

WU

DU,0

SU,0

WR

PU

PR

SU,1

DU,2

SU,2

DU,1

Notes: As in Economy A, a positive productivity shock (λ) leads to higher urban labor demand (DU ,1). The urban-rural wagedifferential is higher than 1, which increases migration. If urban labor supply does not increase as fast as urban labor demand,urban wages reach a higher level. Agglomeration effects (LγU ) further increase urban labor demand (DU ,2). But the urban laborsupply booms as a result of urban natural increase (SU ,2). If urban labor supply increases faster than urban labor demand, urbanwages decrease.

Figure 18: Urban Wages, Migration and Urbanization for both Economies

ECONOMY A ECONOMY B

Time Time Time

WRPR

WUPU

SHARE OF MIGRANTS IN THE URBAN POPULATION

URBAN POPULATION

URBAN WAGE

100

t0 t1 … t0 t1 … t0 t1 …

Notes: This figure shows the evolutions of the urban wage, the share of migrants in the urban population, and urban populationover time for the two economies.

30

TABLE 1: Growth Rates (%) of Selected Cities in Europe and The United States, 1700-1950

Agglomeration Country Long Periods Short Periods1800-1910

1700-1910

1700-1750

1750-1800

1800-1850

1850-1910

New York United States 4.3 3.5 2.2 2.9 4.8 4.0Helsinki Finland 3.3 2.4 1.4 1.4 3.4 3.2Boston United States 3.3 2.5 2.2 1.1 3.6 3.1Manchester England 3.2 2.8 1.6 3.1 2.6 3.7Philadelphia United States 3.0 2.9 3.5 2.0 3.7 2.4Baltimore United States 2.9 . . 2.7 3.8 2.1Belfast Northern Ireland 2.8 2.6 3.1 1.6 3.0 2.6Oslo Norway 2.8 1.9 1.1 0.7 2.2 3.4Zurich Switzerland 2.8 1.6 0.4 0.0 3.1 2.6Budapest Hungary 2.8 2.2 1.0 2.0 2.1 3.4Berlin Germany 2.8 2.0 1.5 0.8 1.9 3.5Munich Germany 2.5 1.5 0.6 0.4 1.8 3.0Glasgow Scotland 2.2 2.0 1.3 2.1 3.2 1.4Brussels Belgium 2.1 1.1 -0.5 0.4 2.3 2.0Liverpool England 2.0 2.3 2.6 2.7 3.1 1.1London England 1.9 1.2 0.3 0.7 1.7 2.0Vienna Austria 1.9 1.4 0.9 0.9 1.4 2.2Birmingham England 1.8 2.1 2.5 2.2 2.4 1.4Hamburg Germany 1.8 1.2 0.5 0.7 0.3 3.1Edinburgh Scotland 1.7 1.3 0.9 0.8 1.7 1.6Rotterdam Netherlands 1.6 0.9 -0.3 0.4 1.1 2.1Marseille France 1.6 0.9 -0.6 0.8 1.3 1.7Barcelona Spain 1.6 1.4 0.8 1.4 1.6 1.7Antwerp Belgium 1.5 0.7 -1.0 0.8 1.1 1.8Paris France 1.5 0.8 0.2 -0.1 1.3 1.7Stockholm Sweden 1.4 1.0 1.1 0.2 0.4 2.2Copenhagen Denmark 1.4 0.9 0.4 0.5 0.4 2.2Lyon France 1.4 0.8 0.3 -0.1 1.0 1.8Geneve Switzerland 1.4 0.9 0.5 0.2 1.2 1.6Rome Italy 1.3 0.7 0.1 -0.1 0.2 2.2Milan Italy 1.3 0.7 -0.1 0.2 0.8 1.7Ghent Belgium 1.1 0.6 -0.7 0.8 1.5 0.7Madrid Spain 1.1 0.9 0.1 1.0 0.3 1.8Amsterdam Netherlands 0.9 0.4 0.1 -0.2 0.3 1.4Dublin Ireland 0.8 1.0 1.5 0.9 0.5 1.0Sevilla Spain 0.5 0.4 -0.2 0.8 0.3 0.6Naples Italy 0.4 0.6 0.8 0.7 -0.1 0.9Lisbon Portugal 0.4 0.3 0.3 0.2 0.2 0.5Chicago United States . . . . . 7.4

Average 1.9 1.4 0.8 1.0 1.7 2.2

Notes: This table shows the average annual urban growth rate (%) for selected periods for selected European and American citiesin 1700-1910. We only include cities from the following countries or regions, for which we have consistent data on urbanizationover time: England (incl. Wales and Northern Ireland), Scotland, Ireland, France, Belgium, The Netherlands, Germany, Austria(incl. Hungary), Switzerland, Italy, Spain, Portugal, Scandinavia and the United States. See Data Appendix for data sources.

31

TABLE 2: Growth Rates (%) of Selected Cities in Africa, 1950-2010 (Decreasing Order)

Agglomeration Country 1950-2010 1950s 1960s 1970s 1980s 1990s 2000s

Yamoussoukro Côte d’Ivoire 12 14.9 15.6 13.1 8.9 9.9 9.8Nouakchott Mauritania 10.4 9.6 22.5 17.6 8.1 2.8 3.2Lilongwe Malawi 10.4 14.9 15.6 13.8 7.9 6 4.5Abuja Nigeria 8.1 1.9 7.6 10 10.2 9.7 9.2Yaoundé Cameroon 7.4 8.9 9.3 8.5 6.5 5.7 5.6Huambo Angola 7.3 9.4 5.1 9.6 7.9 5.9 6Abidjan Côte d’Ivoire 7.2 11.4 11.1 9.7 4.3 3.7 3.2Ouagadougou Burkina Faso 7 6 6.9 8.7 7.3 5.5 7.6Kigali Rwanda 7 6.9 6 8.1 5.5 8.5 6.8Lusaka Zambia 6.9 11.4 11.8 6.7 3.6 3.6 4.8Conakry Guinea 6.9 13.7 13.2 5.4 3.1 3.2 3.5Monrovia Liberia 6.9 17.5 8.1 7.1 12.4 -2.2 -0.3Dar es Salaam Tanzania 6.8 9.2 8.2 8.9 4.6 4.9 4.9Niamey Niger 6.8 9.2 8.3 7.8 4.7 4.6 6Pointe-Noire Congo 6.8 14.5 6.5 6.5 5.3 4 4.1N’Djaména Chad 6.6 12.3 8.3 7.7 3.9 4 4Cotonou Benin 6.5 13.8 8.4 7.5 4.1 2.4 3.2Lomé Togo 6.5 11.2 7.3 6 5 4.9 4.9Kaduna Nigeria 6.4 11 10.4 9 4.3 2.1 2.2Kinshasa DRC 6.4 8.2 9.2 6.7 5.5 4.4 4.5Kananga DRC 6.1 18.9 12 -2.5 0.5 4.4 4.7Luanda Angola 6.1 4.7 7.7 7.7 5 5.2 6.3Lagos Nigeria 6 8.9 6.4 6.2 6.4 4.3 4Port Harcourt Nigeria 5.8 8.4 7 6.1 3.5 4.8 5.2Benin City Nigeria 5.6 5.4 7 7.5 7.5 3.5 3Kano Nigeria 5.6 6.4 9 9.6 4.5 2.2 2.3Douala Cameroon 5.5 4.9 6.9 6.7 5.1 4.7 4.7Khartoum Sudan 5.5 6.6 6.6 5.9 7.3 4 2.6Jos Nigeria 5.4 8.3 8.2 8.1 4.1 2.1 2.2Nairobi Kenya 5.4 7.8 6.2 5 4.8 4.8 3.9Bamako Mali 5.3 3.9 5.5 8.2 4.3 4.4 5.4Mogadishu Somalia 5.2 3.1 11.2 7.3 6.5 1.5 1.7Kisangani DRC 5.2 13.4 5.8 2.2 2.1 3.7 4.3Mbuji-Mayi DRC 5.2 6.9 6.8 4.5 3.5 4.5 4.9Kumasi Ghana 5.1 8.4 4.7 2.6 4.4 5.5 5Brazzaville Congo 5 4.1 6.7 6.5 4.7 3.8 4.3Aba Nigeria 4.9 7.8 6.6 5.9 3.5 2.7 2.9Kampala Uganda 4.8 3.7 9.5 3.3 4.9 3.8 3.8Maiduguri Nigeria 4.8 7.7 7.5 6.8 3.7 1.6 1.7Lubumbashi DRC 4.7 7.3 5.3 4 3 4 4.5Matola Mozambique 4.6 4.5 4.7 4.7 4.6 4.6 4.3Dakar Senegal 4.6 5.7 5.7 4.6 3.9 3.7 3.7Accra Ghana 4.5 8.3 4.8 3.2 3.3 3.4 4Enugu Nigeria 4.4 6.1 4.2 5.6 3.4 3.3 3.6Maputo Mozambique 4.3 7 7.4 4 3.5 2.8 1.1Onitsha Nigeria 4.2 5.7 4.2 2.8 2.7 4.7 5Antananarivo Madagascar 4 3.6 3.7 4.8 5 3.7 3.4Harare Zimbabwe 4 5.7 5.3 4 5.4 2.8 1Vereeniging South Africa 3.9 4.8 5 4 5.1 1.9 2.7Mombasa Kenya 3.9 5.4 4.8 3.3 3.1 3.7 3.2Freetown Sierra Leone 3.9 2.6 5.6 5.8 3.9 2.7 2.8Ogbomosho Nigeria 3.5 6.5 4.3 2.5 2.5 2.5 2.7Addis Ababa Ethiopia 3.4 2.8 3.5 4.9 4.3 2.9 2.1Ilorin Nigeria 3.3 4.6 4.1 3.8 2.8 2.1 2.2Ibadan Nigeria 3.1 2.4 3.6 3.9 3.9 2.5 2.5Durban South Africa 3.1 3.4 2.4 3.6 3.6 3.2 2.2Ekurhuleni South Africa 3 2.2 2.8 2.1 3.3 4.3 3.5Port Elizabeth South Africa 2.9 4.2 5.1 2.1 3.4 1.5 1.4Cape Town South Africa 2.9 2.7 3.3 3.7 3 2.3 2.5Pretoria South Africa 2.8 4.3 3 2 2.8 1.8 3.1Johannesburg South Africa 2.4 2.5 2.3 1.4 1.4 3.7 3.3

Average 5.5 7.5 7.1 6.0 4.7 3.8 3.9

Notes: This table shows the average annual urban growth rate (%) for selected periods for selected African cities in 1950-2010.These cities are agglomerations with 750,000 inhabitants or more in 2010 (WUP 2011). See Data Appendix for data sources.

32

APPENDIX

A-1. DATA CONSTRUCTION [TO BE COMPLETED]

REFERENCES [TO BE COMPLETED]

33

Appendix Figure A.1: Crude Rates of Natural Increase for France, 1853-1910

Notes: This figure plots the crude rate of natural increase (per 1,000 people) for France, rural France, urban France (consideringall cities, including Paris) and Paris in 1853-1910. The crude rate of natural increase is the difference between the crude birthrate and the crude death rate. The data is missing for separate areas before 1853. The average urban crude rate of naturalincrease is 0.1% in 1800-1910 (for available years only). See Data Appendix for data sources.

Appendix Figure A.2: Crude Rates of Natural Increase for Germany, 1811-1912

Notes: This figure plots the crude rate of natural increase (per 1,000 people) for Germany, rural Germany, urban Germany(considering all cities, including Berlin) and Berlin in 1811-1912. The crude rate of natural increase is the difference betweenthe crude birth rate and the crude death rate. The data is missing for all urban and rural areas before 1866. The average urbancrude rate of natural increase is 0.9% in 1800-1910 (for available years only). See Data Appendix for data sources.

34

Appendix Figure A.3: Crude Rates of Natural Increase for the United States, 1825-1910

Notes: This figure plots the crude rate of natural increase (per 1,000 people) for the United States, the rural United States, theurban United States (considering five cities for which we have consistent data: Boston, Baltimore, New York, Philadelphia andChicago) and New York in 1825-1910. The crude rate of natural increase is the difference between the crude birth rate and thecrude death rate. The data is missing for all urban and rural areas. The data is missing for four cities in 1825-1870. The averageurban crude rate of natural increase is 0.5% in 1800-1910 (for available years only). See Data Appendix for data sources.

Appendix Figure A.4: Crude Rates of Natural Increase for Belgium, 1866-1905

Notes: This figure plots the crude rate of natural increase (per 1,000 people) for Belgium, urban Belgium (considering fivecities for which we have consistent data: Brussels, Neighboring Communes of Brussels, Antwerp, Ghent and Liege) and Brusselsin 1866-1905. The crude rate of natural increase is the difference between the crude birth rate and the crude death rate. Theaverage urban crude rate of natural increase is 0.5% in 1800-1910 (for available years only). See Data Appendix for data sources.

35

Appendix Figure A.5: Crude Rates of Natural Increase for the Netherlands, 1815-1909

Notes: This figure plots the crude rate of natural increase (per 1,000 people) for the Netherlands, the urban Netherlands (con-sidering three cities for which we have consistent data: Amsterdam, Rotterdam and The Hague) and Amsterdam in 1815-1909.The crude rate of natural increase is the difference between the crude birth rate and the crude death rate. The data is missing fortwo cities before 1879. The average urban crude rate of natural increase is 0.4% in 1800-1910 (using Amsterdam for availableyears only). See Data Appendix for data sources.

Appendix Figure A.6: Crude Rates of Natural Increase for Sweden, 1800-1910

Notes: This figure plots the crude rate of natural increase (per 1,000 people) for Sweden, urban Sweden (considering all cities),rural Sweden and Stockholm in 1800-1910. The crude rate of natural increase is the difference between the crude birth rate andthe crude death rate. The data is missing for Stockholm in 1800. The average urban crude rate of natural increase is 0.3% in1800-1910 (for available years only). See Data Appendix for data sources.

36

Appendix Figure A.7: Crude Birth Rates and Total Fertility Rates for Africa, 2000

Notes: This figure plots the relationship between the crude birth rate (per 1,000 people) and the total fertility rate (per womanaged 15-49) for the urban areas (considering all cities, including the largest city) and the rural areas of 39 Sub-Saharan Africancountries for the closest year to 2000, depending on the census year or the survey year. The coefficient correlation between thetwo variables is 0.94 at the national level, 0.85 for the urban areas and 0.92 for the rural areas. See Data Appendix for datasources.

Appendix Figure A.8: Crude Death Rates and Child Mortality Rates for Africa, 2000

Notes: This figure plots the relationship between the crude death rate (per 1,000 people) and the child mortality rate (number ofchildren who die by the age of five, per 1,000 live births) at the national level for 47 Sub-Saharan African countries for the year2000. The coefficient correlation between the two variables is 0.95. We have no separate data on the crude death rate for theurban and rural areas of all countries. See Data Appendix for data sources.

37

Appendix Figure A.9: Migration, Natural Increase and Growth of the Largest City, Africa, 1960-2010

Notes: This figure plots the decomposition of the annual growth of the largest city (%) into annual natural increase (%) andannual residual migration (%). Residual migration includes rural-to-urban migration, international migration and urban reclas-sification. If there is some urban reclassification, residual migration is an upper bound to migration. We use historical census andsurvey data that we have collected for 10 representative Sub-Saharan African countries: Burkina-Faso, Central African Republic,Ethiopia, Ghana, Ivory Coast, Kenya, Madagascar, Malawi, Mali and Senegal. See Data Appendix for data sources.

Appendix Figure A.10: Migration, Natural Increase and Urban Growth, Asia, 1960-2010

Notes: This figure plots the decomposition of annual urban growth (%) into annual natural increase (%) and annual residualmigration (%). Residual migration includes rural-to-urban migration, international migration and urban reclassification. If thereis some urban reclassification, residual migration is an upper bound to migration. We use historical census and survey datathat we have collected for 4 representative Asia countries: Bangladesh, India, Pakistan, Sri Lanka for South Asia, Malaysia, thePhilippines and Thailand for Southeast Asia, and China, Japan and South Korea for East Asia. See Data Appendix for data sources.

38

Appendix Figure A.11: Migration, Natural Increase and Urban Growth, MENA, 1960-2010

Notes: This figure plots the decomposition of annual urban growth (%) into annual natural increase (%) and annual “residualmigration” (%). Residual migration includes rural-to-urban migration, international migration and urban reclassification. If thereis some urban reclassification, residual migration is an upper bound to migration. We use historical census and survey data thatwe have collected for 4 representative MENA countries: Egypt, Iran, Jordan and Tunisia. See Data Appendix for data sources.

Appendix Figure A.12: Migration, Natural Increase and Urban Growth, LAC, 1960-2010

Notes: This figure plots the decomposition of annual urban growth (%) into annual natural increase (%) and annual residualmigration (%). Residual migration includes rural-to-urban migration, international migration and urban reclassification. If thereis some urban reclassification, residual migration is an upper bound to migration. We use historical census and survey data thatwe have collected for 8 representative LAC countries: El Salvador, Guatemala, Mexico and Panama for Central America, andChile, Colombia, Ecuador and Peru for South America. See Data Appendix for data sources.

39