African Cities and The Structural Transformation: Evidence from Ghana and Ivory Coast Remi Jedwab Paris School of Economics & LSE ABCDE 2011 Conference, 01 June 2011. 1 / 33
Research Question Introduction Research Question Where do African cities come from? Are African cities different? = Case study on Ghana 1901-2000 and Ivory Coast 1948-1998. 2 / 33
Introduction Motivation Motivation Urbanization in Sub-Saharan Africa: Larger urban population than Northern America / Western Europe Very few cities one century ago Dramatic urban growth after independencies Similar to Industrial Revolution in developed countries Amongst the highest rates of urban change ever registered High urbanization rates More urbanized than India, slightly less than China If lagging in economic development, not in urbanization 3 / 33
Table 1: Urbanization in Developed Countries 1750-1910 vs. Africa 1901-2000. Variable: Urbanization Rate (%) (Pop. 5000 Localities) Level Change (1) (2) (3) (4) (5) (6) (7) Europe/U.S. 1750 1800 1850 1910 1750-1910 1750-1850 1850-1910 England 15 23 45 75 60 30 30 France 12 12 19 38 26 7 19 Germany 11 9 15 49 38 4 34 United States 5 5 14 42 37 9 28 Africa 1901 1948 1975 2000 1948-2000 1948-1975 1975-2000 Africa 5.5 10.5 21.4 35* 24.5 10.9 13.6 Ghana 5 10.9 31.3 43.8 32.9 20.4 12.5 Ivory Coast 0 5.7 34.6 56.8 51.1 28.9 22.2 Note: the urbanization rate is defined as the total population in localities with more than 5000 inhabitants over total population times 100. The main source of historical urbanization data for developed countries and Africa is Bairoch (1988). Data from Ghana and Ivory Coast comes from my sample. * No such figure is available for Africa in 2000, so I take the projected estimate by Bairoch.
Introduction Motivation Motivation Urbanization and Structural Transformation: Development associated with the structural transformation: decline of agriculture, rise and fall of manufacturing, rise of services. Structural transformation in developed countries, China, India: Rising agricultural productivity releases labor for the modern sector High productivity in manufacturing/tradable services exports Structural transformation in Africa: Low yields + uncompetitive manufacturing and service sectors. But dramatic urban change. 4 / 33
Introduction Motivation Figure: Manufacturing/Services and Urbanization in Developing Countries, 2000. 5 / 33
Introduction Motivation Figure: Primary Exports and Urbanization in Africa, 2000. 6 / 33
Introduction Motivation Figure: Primary Exports and Cities in West Africa, 2000. 7 / 33
This Paper s Approach Introduction This Paper s Approach Develop a new structural transformation model where primary exports push urbanization via consumption linkages. Assemble a historical data set on cities in Ghana/Ivory Coast. We combine this with a district panel data set on cocoa-coffee production. Ghana 1900-2000: N = 79x9, Ivory Coast 1948-98: N = 46x6. Test if cash crop production drives urbanization. Identification strategy: cocoa produced by consuming the forest. When cocoa trees are too old (25 years), no choice but to deforest a new region. Regional cycles as the cocoa frontier moves. Investigate the employment structure of those consumption cities. 8 / 33
Findings Introduction Findings Large positive effect of cash crop production on local urbanization. It respectively explains 54.3% and 90.4% of urban growth in non-national cities in Ghana and Ivory Coast. Lower magnitude in Ghana due to higher state taxation (diversion of urbanization effects) + mining. Strong evidence for consumption linkages. Cocoa farmers spend their rising income on urban goods and services. Consumption cities consist mostly of landowners and service workers. Very small manufacturing sector and no economic diversity. 9 / 33
Contributions Introduction Contributions 1 Structural transformation: Matsuyama 1992, Gollin & Rogerson 2010, Michaels et al. 2011. 2 Urbanization and growth in developing countries: Fay & Opal 2000, Henderson 2003, Duraton 2008, Venables 2010. 3 Cash crop windfalls and resource curse: Bevan, Collier & Gunning 1987, Collier & Goderis 2009, Michaels 2010. 4 Geography versus history in development: Davis & Weinstein 2002, Glaeser & Gottlieb 2009, Nunn & Qian 2010. 10 / 33
Road Map Introduction Road map 1 Conceptual framework 2 Background and data 3 Empirical strategy 4 Channels 5 Discussion on consumption cities 6 Concluding comments 11 / 33
Conceptual Framework Conceptual Framework Main Model Main Model: Small open economy. Agricultural productivity such that everyone lives at the subsistence level. Natural resource boom: the owners of the natural resource exchange it against imported manufactured goods or food. Non-homothetic preferences (Engel curve) surplus spent on: imported manufactured goods (traders paid by imported food) non-tradable services (service workers paid by imported food) The urban sector grows, but consumption cities (non-tradable services) 12 / 33
Conceptual Framework Conceptual Framework Comments and Extensions Comments and Extensions: State taxation: the location of urbanization effects depends on the distribution of ownership rights over the natural resource. Model of non-industrialization, and not a model of deindustrialization (see the Dutch Disease literature). Assume learning-by-doing effects are lower in non-tradable services than in manufacturing. A country with a comparative advantage in natural resources has more cities, but those cities grow less in the long run. 13 / 33
Background and Data Background Background on Cocoa, Ghana and Ivory Coast Largest producers in the world. Also coffee in Ivory Coast. Production take-off in the 1920s in Ghana, 1960s in Ivory Coast. High contribution to exports (60%) and GDP (15-20%). Produced by consuming the forest: Cocoa farmers go to a patch of virgin forest and plant cocoa trees. Pod production peaks after 10 years, and declines thereafter. When cocoa trees are too old (25 years), new cycle in a new forest. 14 / 33
Background and Data Cocoa Production Data Cocoa Production Data Variables: C dt is cocoa production value in district d between t 1 and t. Value (in 2000$): volume deflated producer price. t Ivory = 1948, 1955, 1965, 1975, 1989, 1998 t Ghana = 1901, 1911, 1921, 1931, 1948, 1960, 1970, 1984, 2000 Ghana: cocoa districts and not administrative districts. Ivory Coast: district-year coffee production data as well. Primary Sources: Government agency responsible for cocoa production and taxation: CAISTAB in Ivory Coast, COCOBOD in Ghana. 15 / 33
Urban Data Background and Data Urban Data Variables: U dt is the urban population of district d at time t. Urban: cities are defined as 5000 localities. t Ivory = 1901, 1911, 1921, 1931, 1948, 1955, 1965, 1975, 1989, 1998 t Ghana = 1901, 1911, 1921, 1931, 1948, 1960, 1970, 1984, 2000 Primary Sources: Censuses + administrative counts: size for each city-year. Use GIS to extract urban population for any district decomposition. 16 / 33
Background and Data Mapping Figure: Land Suitability to Cocoa Cultivation and Administrative Boundaries. 17 / 33
Background and Data Mapping Figure: Value of Cash Crop Production (1900-2000) and Cities (2000). 18 / 33
Background and Data Mapping Figure: District Density of Cocoa Production and Cities in 1948. 1948 in Ghana, 1948 in Ivory Coast. 19 / 33
Background and Data Mapping Figure: District Density of Cocoa Production and Cities in 1960-1965. 1960 in Ghana, 1965 in Ivory Coast. 20 / 33
Background and Data Mapping Figure: District Density of Cocoa Production and Cities in 1970-1975. 1970 in Ghana, 1975 in Ivory Coast. 21 / 33
Background and Data Mapping Figure: District Density of Cocoa Production and Cities in 1984-1988. 1984 in Ghana, 1988 in Ivory Coast. 22 / 33
Background and Data Mapping Figure: District Density of Cocoa Production and Cities in 1998-2000. 2000 in Ghana, 1998 in Ivory Coast. 23 / 33
Background and Data Mapping Figure: District Density of Cocoa Production and Cities in 2009. 2009 in Ghana, 2009 in Ivory Coast. 24 / 33
Empirical Strategy Strategy 1: Long-Term Model Long-Term Model For district d, we run: U d = α + δ C d + ηs d + ɛ d (1) U d : urban growth (inhabitants) of district d between t 0 and T. C d : cocoa+coffee production value of district d between t 0 and T. S d : district-level controls. Instrument cash crop production with a district dummy for being highly suitable to cocoa cultivation (also try spatial RDD). 25 / 33
Set of Controls Empirical Strategy Long-Term Model 1 Political Economy: district dummies for having a national city (capital and second largest), regional capital. 2 Economic geography: district area (sq km), district dummies for having a paved road (1965), railway (1965), international port (1965), Euclidean distance (km) to the coast. 3 Physical geography: 1900-2006 district average annual precipitations (mm), average annual maximal temperature ( C). 26 / 33
Table 2: Cash Crop Production and Urbanization, Long-Term Model. Dependent Variable: Panel A: Main Equation District Urban Growth (Number of New Urban Inhabitants, Between t 0 and T ) Ivory Coast, 1948-1998 Ghana, 1901-2000 OLS OLS 2SLS 2SLS OLS OLS 2SLS 2SLS (1) (2) (3) (4) (5) (6) (7) (8) District Value of Cash Crop Production 133.6** 79.5*** 71.0*** 117.1*** 63.7*** 93.7*** 63.5*** 92.7** (Millions of 2000$, Between t 0 and T ) [60.2] [26.5] [20.9] [45.3] [16.1] [27.0] [19.0] [41.8] Panel B: First Stage High Suitability Dummy 1,307*** 1,083*** 879.0*** 627.5*** [258] [240] [108.7] [107.6] Kleibergen-Paap rk Wald F Stat 23.6 15.1 62.1 28.8 Observations 50 50 50 50 79 79 79 79 R-squared 0.50 0.88 0.48 0.88 0.85 0.89 0.85 0.89 National Capital Dummy, Area Y Y Y Y Y Y Y Y Controls N Y N Y N Y N Y Note: Standard errors in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%. National Capital Dummy is a dummy equal to one if the district contains Abidjan, Bouaké or Yamoussoukro in Ivory Coast, and Accra or Kumasi in Ghana. I include various controls. Political economy: a dummy equal to one if the district contains a regional capital. Economic geography: dummies for whether the district has a paved road, a railway or an international port in 1965, as well as Euclidean distance (km) to the coast. Physical geography: 1900-2006 average annual precipitations (mm) and average maximal temperature ( C).
Empirical Strategy Strategy 2: Short-Term Model Short-Term Model For district d and year t, we run: U d,t = α d + β t + δ C d,t + η ts d + ɛ d,t (2) U d,t : urban growth (inhabitants) of district d between t 1 and t. C d : cocoa+coffee production value of district d between t 1 and t. S d : district-level controls, whose effects η t are time-varying. Instrument: high suitability dummy x distance to Eastern border x time trend (controlling for distance to Eastern border x time trend). 28 / 33
Table 3: Cash Crop Production and Urbanization, Short-Term Model. Dependent Variable: Panel A: Main Equation District Urban Growth (Number of New Urban Inhabitants, Between t and t 1) Ivory Coast, 1948-1998 Ghana, 1901-2000 OLS OLS 2SLS 2SLS OLS OLS 2SLS 2SLS (1) (2) (3) (4) (5) (6) (7) (8) District Value of Cash Crop Production 101.1*** 72.4*** 89.1*** 63.2* 25.0 22.6 167.8* 57.3* (Millions of 2000$, Between t and t 1) [29.4] [18.2] [34.5] [36.3] [15.6] [20.7] [95.9] [33.3] Distance to Border x [Year - t 0 ] -0.2-1.9-0.4 0.0 [1.4] [2.4] [0.8] [0.0] Panel B: First Stage High Suitability Dummy x Dist. to Border x [Year - t 0 ] 0.03*** 0.02*** 0.01*** 0.01*** [0.01] [0.00] [0.00] [0.00] Distance to Border x [Year - t 0 ] -0.01-0.01* 0.00*** 0.00 [0.001] [0.00] [0.00] [0.00] Kleibergen-Paap rk Wald F Stat 18.1 11.8 33.6 36.1 Observations 250 250 250 250 632 632 632 632 R-squared 0.83 0.94 0.84 0.68 0.75 0.76 0.74 0.38 District and Year Fixed Effects Y Y Y Y Y Y Y Y National Capital Dummy, Area x Year Y Y Y Y Y Y Y Y Baseline Controls x Year N Y N Y N Y N Y Note: Standard errors in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%. National Capital Dummy is a dummy equal to one if the district contains Abidjan, Bouaké or Yamoussoukro in Ivory Coast, and Accra or Kumasi in Ghana. It is interacted with a time trend. I also include various baseline controls interacted with a time trend. Political economy: a dummy equal to one if the district contains a regional capital. Economic geography: dummies for whether the district has a paved road, a railway or an international port in 1965, as well as Euclidean distance (km) to the coast. Physical geography: 1900-2006 average annual precipitations (mm) and average maximal temperature ( C).
Empirical Strategy Comments, Robustness and Additional Results Comments, Robustness and Additional Results The long-term effect respectively explains 54.3% and 90.4% of urban growth in non-national cities in Ghana and Ivory Coast. Lower magnitude in Ghana due to higher state taxation (diversion of urbanization effects) + mining. No significant effect of cocoa production on rural growth. Urban growth effect equally decomposed between existing cities growing further and new cities. Cities keep growing in old producing regions: cocoa production launches a self-reinforcing urbanization process. Infrastructure investments and demographic growth account for urban resilience. 30 / 33
Channels Channels Centre-Ouest region of Ivory Coast (LSMS 1985-88, EP 2002). 460,000 new urban inhabitants in 1988-1998: Demand side of urban labor. Cocoa-producing households: represent 76.5% of population growth 30% wealthier than non-cocoa farmers within the same village spend 57.9% of urban goods and services Boom in urban expenditure Supply side of labor. Decomposition of urban employment growth: primary sector = +56.5% (cocoa +34.4%) secondary sector = -9.1% tertiary sector = +52.6% (trade +24.3%, personal services +17.6%, transport and communications +8.8%, education and health +2.7%, business services and banking +2%). 31 / 33
Discussion on Consumption Cities & Research Agenda Discussion on Consumption Cities & Research Agenda Ghanaian and Ivorian cities consist of landowners, traders and service workers. Abidjan/Accra: 3/4 in the tertiary sector. Small manufacturing sector and less economic diversity even in regions that experienced a cocoa boom several decades ago. Cities in the old cocoa-producing regions do no not collapse, on the contrary. Evidence on decreasing income when cocoa leaves? 32 / 33
Concluding Comments Concluding Comments Cash crop exports explain more than half of local urbanization in both countries. This does not include the effect of cash crops on the growth of national cities. But this effect gives consumption cities of landowners, traders and service workers. Implications? Growth effect of consumption cities vs. production cities? Urban resilience in old cocoa-producing regions. But scenario of demographic growth + resource exhaustion. 33 / 33