The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Nga Thi Viet Nguyen and Felipe F. Dizon Acknowledgements This study was prepared by Nga Thi Viet Nguyen and Felipe F. Dizon. Additional contributions were made by Brian Blankespoor, Michael Norton, and Irvin Rojas. Marina Tolchinsky provided valuable research assistance. Administrative support by Siele Shifferaw Ketema is gratefully acknowledged. Overall guidance for this report was received from Andrew L. Dabalen. Joanne Gaskell, Ayah Mahgoub, and Aly Sanoh provided detailed and careful peer review comments. The team greatly benefited from the valuable support and feedback of Félicien Accrombessy, Prosper R. Backiny-Yetna, Roy Katayama, Rose Mungai, and Kané Youssouf. The team also thanks Erick Herman Abiassi, Kathleen Beegle, Benjamin Billard, Luc Christiaensen, Quy-Toan Do, Kristen Himelein, Johannes Hoogeveen, Aparajita Goyal, Jacques Morisset, Elisée Ouedraogo, and Ashesh Prasann for their discussion and comments. 2 Abbreviations and Acronyms AIDS Acquired Immune Deficiency Syndrome CGIAR Consultative Group for International Agricultural Research CMU Country Management Unit ECOWAS Economic Community of West African States EMC Continuous Multi-Sectoral Survey (Enquête Multisectorielle Continue) EMICOV Integrated Modular Household Well-Being Survey (Enquête Modulaire Intégrée sur les Conditions de Vie des Ménages) ENV Household Living Standards Survey (Enquête sur le Niveau de Vie des Ménages) FAO Food and Agriculture Organization FEWSNET Famine Early Warning Systems Network GADM Global Administrative Area Database GDP Gross Domestic Product HI Herfindahl Index HIV Human Immunodeficiency Virus ICT Information and Communication Technology IFPRI International Food Policy Research Institute LMI Low and Middle Income NEG New Economic Geography OECD Organisation for Economic Cooperation and Development PAD Project Appraisal Document PFR Rural Land Use Plan (Plan Foncier Rural) PID Project Information Document PPP Purchasing Power Parity QUIBB Basic Well-Being Indicator Questionnaire (Questionnaire des Indicateurs de Base du Bien-être) R&D Research and Development SCD Systematic Country Diagnostic SHIP Survey-Based Harmonized Indicators Program SMS Short Message Service SSA Sub-Saharan Africa TFP Total Factor Productivity UN United Nations WAEMU West African Economic and Monetary Union WDI World Development Indicator 3 Table of Contents Acknowledgements ....................................................................................................................................... 2 Abbreviations and Acronyms ....................................................................................................................... 3 Executive Summary ...................................................................................................................................... 7 Chapter 1: Location and Prosperity ............................................................................................................ 11 Motivation and Objectives ...................................................................................................................... 11 Regional Context .................................................................................................................................... 12 Data ......................................................................................................................................................... 14 Chapter 2: Geography of Welfare – Three Building Blocks....................................................................... 17 Natural Endowment ................................................................................................................................ 17 Agglomeration Economies...................................................................................................................... 20 Market Access......................................................................................................................................... 22 Chapter 3: Spatial Disparities in Welfare and Poverty ............................................................................... 25 Leading and Lagging Areas .................................................................................................................... 26 Poverty Rates, Poverty Mass, and Poverty Density ................................................................................ 30 Access to Services................................................................................................................................... 33 Quality of Life and Characteristics of Poor People ................................................................................ 36 Chapter 4: Geographical Differences in Agricultural Activity ................................................................... 42 Employment in Agriculture .................................................................................................................... 44 Agricultural Productivity ........................................................................................................................ 45 Assets, Inputs, and Output Markets ........................................................................................................ 53 Chapter 5: Putting It All Together .............................................................................................................. 59 Relationship between Welfare and Agricultural Productivity ................................................................ 59 Three Sets of Explanatory Variables for Three Building Blocks ........................................................... 60 Correlates of Welfare .............................................................................................................................. 61 Correlates of Agricultural Productivity................................................................................................... 63 Chapter 6: Policy Recommendations and Further Studies .......................................................................... 66 Urbanization............................................................................................................................................ 66 Agricultural Productivity ........................................................................................................................ 67 Fiscal Transfers ....................................................................................................................................... 69 Safety Net Programs ............................................................................................................................... 69 Limitations and Further Studies .............................................................................................................. 70 References ................................................................................................................................................... 73 Appendix A: Agro-ecological Zones .......................................................................................................... 79 4 Appendix B: Market Accessibility Index – Methodology .......................................................................... 82 Appendix C: Extra Materials ...................................................................................................................... 84 Appendix D: Summary of Findings on Agricultural Activities across Countries, by Zone ....................... 93 Appendix E: Agricultural Data – Notes on Model Construction ................................................................ 94 Maps, Tables, Figures, and Boxes Map 1.1: Climate classification ................................................................................................................................... 13 Map 2.1: Abundant in terms of precipitation levels .................................................................................................... 18 Map 2.2: Agro-ecological zones, by country ............................................................................................................... 19 Map 2.3: Concentration of agglomeration economies in the South ............................................................................. 22 Map 2.4: Road networks .............................................................................................................................................. 23 Map 2.5: Concentration of high market access in the South, around capitals and economic capitals ......................... 24 Map 3.1: Cluster of leading areas in the South, around the capitals and the economic capitals, or along country borders ......................................................................................................................................................................... 27 Map 3.2: In three out of four countries, the North is remarkably poorer than the South. ............................................ 30 Map 3.3: Lower poverty rates in areas around the capitals and the economic capitals, and along the country border 31 Map 3.4: High poverty density in and around the capitals and the economic capitals, and in the South .................... 33 Map 3.5: High variation in access to public services across space (except cellphone coverage) ................................ 35 Map 3.6: Low service coverage in areas with high poverty incidence ........................................................................ 36 Map 3.7: Higher diversity in food consumption basket and lower food share from own production in the South ..... 37 Map 3.8: Variation in key food consumed across space .............................................................................................. 38 Map 4.1: Employment in agriculture relative to other sectors ..................................................................................... 44 Map 4.2: Cash crops across agro-ecological zones...................................................................................................... 49 Map 4.3: Cash crops across Côte d’Ivoire’s agro-ecological zones ............................................................................ 50 Map 4.4: Maize yields across agro-ecological zones ................................................................................................... 51 Map 4.5: Cash crop yields across agro-ecological zones............................................................................................. 53 Map 4.6: Use of inputs and farm land across agro-ecological zones ........................................................................... 54 Map 4.7: Land tenure security across agro-ecological zones ...................................................................................... 57 Map 4.8: Sale of agricultural produce ......................................................................................................................... 58 Table 1.1: Economic per capita growth by international standards ............................................................................. 14 Table 1.2: Geographical data sources .......................................................................................................................... 16 Table 3.1: Better housing conditions for poor households in urban areas or in favorable agro-ecological zones ....... 39 Table 3.2: Fewer family members, lower dependency rates, more likely to be female-headed households, and less likely to have no education among poor households in urban areas and favorable zones ........................................... 40 Table 5.1: Statistical summary .................................................................................................................................... 61 Table 5.2: Factors associated with spatial differences in poverty: coastal location, population density, and market access ........................................................................................................................................................................... 62 Table 5.3: Role of geographical differences in agricultural productivity, natural endowments (temperature, latitude, elevation, coastal location), and spending on fertilizer ................................................................................................ 64 Figure 2.1: Higher share of population living in urban areas but slower urban population growth by African standards ...................................................................................................................................................................... 21 Figure 3.1: Large wealth gap between leading and lagging areas ............................................................................... 27 Figure 3.2: Many leading areas in low-density locations ............................................................................................ 28 Figure 3.3: Low market access in many leading areas ................................................................................................ 29 Figure 3.4: Majority of the poor living in low-density areas ....................................................................................... 32 5 Figure 3.5: Large gaps in public service coverage between the most sparsely and most densely populated areas ..... 34 Figure 4.1: Percentage of population engaged in agriculture, by poor and non-poor .................................................. 42 Figure 4.2: Crops grown by poor and non-poor .......................................................................................................... 47 Figure 5.1: Correlation between poverty and agricultural productivity ....................................................................... 60 Box 4.1: Summary of the agriculture sector based on various world bank documents (Project Information Document [PID], Project Appraisal Document [PAD], and Systematic Country Diagnostic [SCD]) .......................................... 46 Box 4.2: Background on land reform in Benin, Côte d’Ivoire, and Burkina Faso ...................................................... 56 6 Executive Summary West Africa is at the heart of Africa’s transformation. With a gross domestic product (GDP) gro wth rate of more than 5 percent annually, it is the fastest growing region in the continent. Yet, poverty rates remain high, even by African standards. In the four West African countries covered in this report, namely Benin, Burkina Faso, Côte d’Ivoire, and Togo, nearly half of the population lives on less than US$1.90 a day at 2011 purchasing power parity (PPP). This means over 25 million people live in extreme poverty. How is it possible that the sub-region’s high economic growth cannot translate into higher levels of prosperity? The answer lies in where one looks as national averages often mask large disparities at subnational levels. Recent literature on the new economic geography suggests that within-country disparities may be a natural outcome of the development process. As a country develops, economic activity clusters in regions endowed with more favorable agro-ecological conditions, more abundant natural resources deposits, or simply a better location. More economic opportunities in turn attract more people in search of jobs, which consequently increases population density in one area over another. Arguably, the higher concentration of people and economic activity leads to economies of scale. Such benefits can be further enhanced with the existence of market access for products, labor, and ideas, which continues to boost these regions’ income and attractiveness to people and firms. This virtuous cycle of development makes it difficult for poor regions to catch up. This report aims to assess the spatial disparities in economic development along four important dimensions: (a) It provides stylized facts of the underlying forces behind within-country inequality, namely natural endowment, agglomeration economies, and market access. These are the three building blocks of the economic geography literature; (b) It examines spatial disparities in welfare and poverty. As the agricultural sector is a cornerstone of the economy in this sub-region, the report explores geographical differences in agricultural activity; (c) It quantifies the roles of natural endowment, agglomeration economies, and market access in determining the spatial distribution of welfare and agricultural productivity; (d) It suggests a number of policy guidelines that may help improve shared prosperity across space. However, we acknowledge that since poverty is a multidimensional concept, many other factors could potentially contribute to the observed within-country inequality, yet are not covered by the scope of this study. These may include social elements such as ethnicity, nutrition, and health, economic conditions such as prices and markets, and political dimension such as institutions and conflict. A Tale of Two Regions In terms of agro-ecological endowment, two distinct groups emerge within the sub-region. Being the most northerly and the only landlocked country in the set, Burkina Faso has a noticeably different agroecosystem, being generally drier and less fertile. It also has the most dispersed population in the sub-region. The rest— Benin, Côte d’Ivoire, and Togo—are coastal and located at the same latitude and therefore share similar agro-ecological endowments. Of the four countries, Togo has the highest population density. Within each of the three coastal countries, there also exists a tale of two regions, with the South having more favorable condition for agricultural activities, access to the sea, and higher population density than the North. Thus, the pattern of market access is similar to that for natural endowment and agglomeration, with higher levels of market access concentrating in the South. The picture is slightly different in landlocked Burkina Faso. While it shares a similar geographical pattern of agro-ecological characteristics as the one found in the neighboring countries (i.e., North vs. South), its 7 population and market access concentrate only in the Central region, home of the two largest cities: Ouagadougou and Bobo-Dioulasso. A Tale of Two Economies In the three coastal countries, the North is markedly poorer and has a larger share of population employed in the agriculture sector than the South. In Togo particularly, poverty in the far North may be more than three times as high as in the far South. However, the pattern of poverty is reversed in Burkina Faso, in part because of the possession of livestock among northern residents. The spatial distribution of crops grown also varies between North and South. Cash crops, or crops produced for commercial value, are more prevalent in the South (except cotton) in terms of the proportion of farmers growing them. However, the geographical coverage of cotton production is quite different and resembles a belt covering the southern parts of Burkina Faso and the northern parts of Benin, Côte d’Ivoire, and Togo. Interestingly, this cotton belt overlaps with areas of higher agglomeration distant from the capital city of each country. Looking beyond monetary poverty and agricultural activity, the quality of life of the poor as measured by the extent of food intake diversification, access to basic services, and housing conditions increases significantly from North to South. In other words, given two poor individuals with similar incomes, the one living in the South enjoys a more diverse food basket, has a higher chance of having access to electricity and sanitation, and has a higher probability of living in a house with either a concrete roof or brick walls than his fellow citizen in the North. These patterns are consistent across all four countries. At subnational levels, differences in welfare are even more pronounced between leading and lagging locations.1 A location is defined as “leading” if per capita consumption for an average household living there is higher than the national average. A typical household in a leading area may consume as much as seven times more overall than a similar household in a lagging area. This gap is highest in Benin and lowest in Burkina Faso. Notably, many of the leading areas have not yet maximized the benefits of agglomeration economies. This observation is especially clear in Burkina Faso and Côte d’Ivoire, where approximately half of the leading regions are located in low-density areas. This suggests scope for greater concentration of economic activities and labor in these locations in order to further take advantage of economies of scale and boost economic development. Moreover, in each country, there exists geographical pockets of poverty that may be resistant to policy- induced changes. These lagging areas are characterized by a combination of high poverty rates and a low number of poor people per square kilometer. As a result, the unit cost of a poverty-targeted program may be extremely high in these areas. Given budget constraints, the government may not be able to reach this population group. Spatial Disparities Explained There has been a long ongoing debate in the economic geography literature on whether a location’s levels of per capita income and other economic dimensions are determined by geographical and ecological variables. Many researchers have provided evidence supporting the view that such links are strong, while others have argued that the role of geography in explaining spatial patterns of per capita income operates through various direct channels (e.g., productivity and trade) or indirect channels (e.g., choice of political and economic institutions), with little direct effect of geography on incomes. How therefore does this play out in the case of Benin, Burkina Faso, Côte d’Ivoire, and Togo? 1 Subnational levels consist of communes in Benin, provinces in Burkina Faso, departments in Côte d’Ivoire, and prefectures in Togo. 8 As it turns out, except for being coastal or landlocked, agro-ecological characteristics do not appear to be directly associated with a location’s per capita income. The relationship between geography and welfare is indeed being mediated by agglomeration economies and market access. In other words, locations favorable to growth will either attract people or experience stronger population growth and at the same time receive investments in infrastructure. Thus, when controlling for population density and market access, the correlation between welfare and geographical variables (except for being located along the coast) is no longer significant. If natural endowment plays any role in explaining spatial disparities in welfare, the key factor is location near the coast. Given two locations with the exact same population density and market access, the one on the coast is 21 percent richer than the one located inland. The fact that the economic benefits of coastlines remain strong (i.e., they have not been arbitraged away by migration or increased market access) reveals the untapped potential for economic development provided by access to international trade for the three coastal countries (Benin, Côte d’Ivoire, and Togo). However, the story is quite different when looking at agricultural productivity measured by maize yields. This report focuses on maize yields because this food crop is fairly prevalent across all four countries and is found in most areas in each country. This allows for some degree of comparability for yields across zones and across countries. What is notable is the persistence of the correlation between geography and agricultural productivity regardless of whether population density, market access, or farm inputs are taken into account. In contrast to the new economic geography literature suggesting that agglomeration economies and market access can help farmers take advantage of better prices, a wider selection of agricultural inputs, and better markets for harvested crops, this link is weak in the sub-region. This finding implies that there may in fact be two types of agriculture: a subsistence agriculture, whereby most crops are cultivated for home consumption and where investments are less sensitive to market access, and a commercial agriculture, which might concentrate along coastlines and benefit from higher investment in inputs. What Can Be Done? Within-country disparities can be a potential source of tensions between lagging and leading locations and may affect the country’s overall growth and political stability. How can our findings help policy makers reduce geographical differences in welfare while boosting growth? Based on our analysis, we propose four broad policy recommendations: (a) Urbanization: We find that many of the leading areas have not yet maximized the benefits of agglomeration economies, especially in Burkina Faso and Côte d’Ivoire. Based on the new economic geography literature, there is scope for increasing concentration in economic activities and labor in these areas to further take advantage of economies of scale and boost economic development. However, it is important to consider complementary policies to urbanization, including removing barriers to labor mobility so that people can migrate to leading areas, where demand for labor and productivity are higher, and investing in urban infrastructure and the provision of public services to accommodate a potentially larger number of users. (b) Increasing agricultural productivity: Not all rural families can move to urban locations. For those staying in the agriculture sector in rural areas, policy makers should consider improving welfare by increasing agricultural productivity. Potential areas of improvement include land tenure, irrigation, use of farm inputs such as fertilizer, and research and development. (c) Fiscal budget transfers: Geographical pockets of poverty exist where the costs of reaching the poor are very high. These areas are characterized by a combination of high poverty rates and low poverty density. Another set of lagging areas with little prospect of growth consists of those with unfavorable agro-ecological characteristics and limited opportunities to diversify into non- agriculture sectors. Our quantitative analyses show a persistent link between agro-ecological 9 endowment and agricultural productivity regardless of whether agglomeration, market access, or farm inputs are taken into account. Our findings imply that some lagging areas may not be able to improve their welfare after all. This may call for pro-poor fiscal transfers through a system of inter- region transfers to ensure equity across leading and lagging areas. (d) Safety net programs: Not all poor people, especially the vulnerable, can benefit from the policies proposed above. Thus, the need to maintain strong safety net programs targeting the poor and vulnerable remains strong. New technologies such as e-vouchers and mobile transfers make it possible for such programs to reach targeted beneficiaries in low-density areas in a cost-effective way. Moreover, safety net programs should be part of an overarching poverty reduction strategy consisting of interacting with and working alongside urban policy, agricultural productivity boosting programs, and other policies aimed at eradicating poverty and reducing vulnerability. 10 Chapter 1: Location and Prosperity Motivation and Objectives Location is the most critical predictor of a person’s welfare (World Bank, 2009). As of today, a child born in Togo is expected to live nearly 20 years less than a child born in the United States. Moreover, the child will earn a tiny fraction—less than 3 percent—of what her American counterpart will earn (World Bank, 2017). Such disparities in income and living standards within a country are just as unsettling . An urban inhabitant in Togo’s capital, Lomé, has a 16 percent chance of being poor, and a 90 percent chance of having access to electricity. However, these probabilities are reversed for a person from a rural district in far northern Oti prefecture, where residents have an 80 percent chance of falling into poverty and a mere 13 percent chance of having access to electricity. As pointed out in the World Development Report – Reshaping Economic Geography (World Bank, 2009), such within-country disparities can pose a major challenge for policy makers as they present a potential source of increasing tensions between poorer and richer areas. Moreover, if these spatial inequalities persist or widen, they can potentially affect a country’s future growth and political stability. Spatial differences in economic development have long been the subject of study, with a history dating back to the 4th century B.C. and expanding after World War II due to uneven post-war economic recovery and development. Until the 1980s, the study of economic geography was under scrutiny because it undermines the notion of equal opportunity among individuals. However, over the past decades, the field has regained attention in mainstream development debates thanks to new theories of on economic growth and empirical research in this field (Hausmann, 2001). The recent literature on the new economic geography (NEG) implies that within-country disparities may be a natural outcome of the development process, and once established, can be persistent and insensitive to policy-induced changes (see Fujita, Krugman, and Venables, 1999; Puga, 1999; Fujita and Thisse, 2002; World Bank, 2009). Thus, from a policy maker’s perspective, the success of any government policy aimed at improving shared prosperity across locations crucially depends on what drives the observed spatial inequality. This study covers four countries in West Africa under the Country Management Unit (CMU) AFCF2, namely Benin, Burkina Faso, Côte d’Ivoire, and Togo, and makes use of four recently collected household consumption surveys. Given the data limitations, we focus on a static analysis of economic geography in this sub-region. While our discussion is centered around within-country inequalities as these are more relevant to each respective government, we touch upon some aspects of cross-country differences in order to provide a regional context. We also emphasize that our analysis focuses mainly on the spatial distribution of welfare and poverty. To understand the driving forces behind differences of welfare and poverty across space, we base our analysis on the NEG literature and highlight its three building blocks: natural endowment, agglomeration economies, and market access. In addition, we explore the spatial distribution of several key elements that intertwine with poverty, including economic activity, agricultural productivity, household demographics, and access to services (Banerjee and Duflo, 2007). We acknowledge that since poverty is a multidimensional concept, many other factors could potentially contribute to observed within-country inequality, yet are not covered by the scope of this study. These may include social elements such as ethnicity, nutrition, and health, economic conditions such as prices and markets, and political dimension such as institutions and conflict. Our objectives are: (a) To provide stylized facts relevant to the three building blocks of the economic geography literature: natural endowment, agglomeration economies, and market access; 11 (b) To examine static spatial disparities in welfare and poverty together with relevant development indicators such as poor household demographics, access to services, economic activity, and agricultural productivity; (c) To quantify the roles of natural endowment, market access, and agglomeration economies in determining the spatial distribution of welfare; (d) To suggest a number of policy guidelines that may help improve shared prosperity across space. The report is organized as follows. The rest of Chapter 1 provides a glance at the regional context (i.e., where these countries stand in the global economy) as well as an overview of the data used. Chapter 2 introduces the three building blocks of the economic geography literature: natural endowment, agglomeration economies, and market access. Chapter 3 presents stylized facts about the sub-region’s spatial disparities, focusing on welfare, poverty, access to services, and profiles of the poor. Chapter 4 provides details of the geographical distribution of agricultural activity given the important role it plays in the sub-region. Chapter 5 uses the NEG framework to explore correlates of the observed inequalities across space. Finally, Chapter 6 concludes with a policy discussion. Regional Context The majority of the world’s extreme poor, those living on less than US$1.90 a day at 2011 purchasing power parity (PPP), are concentrated in Sub-Saharan Africa (SSA). Within the continent, West Africa is home to some of the poorest nations, where approximately half of the population lives in poverty, and over three quarters of the population have no access to improved sanitation (World Bank, 2017). Along key factors such as institutional quality, labor productivity, and human capital, the region’s geographical characteristics are often considered, to be a key constraint on its economic development. The four West African countries covered in this report (Benin, Burkina Faso, Côte d’Ivoire, and Togo) lie mostly in the tropical savannah climate area (Map 1.1), a common geographic disadvantage identified among countries lagging behind in economic development. Hausmann (2001) shows that, on average, annual economic growth rates in tropical nations are between one-half and a full percentage point lower than in temperate countries. In addition, countries located in tropical areas often show more skewed income distribution and poorer health conditions than their non-tropical counterparts. 12 Map 1.1: Climate classification Source: Kottek et al., 2006. As shown in Table 1.1, these four West African countries are poor by international standards . Not only is their per capita GDP lower than the average of Low and Middle Income (LMI) countries, it is also lower than African averages. Even though these economies, especially Burkina Faso and Côte d’Ivoire, have grown at the impressive rate of approximately 5 percent per year, annual growth in per capita GDP still falls behind those of their peers, in part because of relatively fast population growth. A related fact is the high levels of population density across all four countries. Within this sub-region, Togo is the densest, with the number of people per square kilometer being about three times higher SSA averages, and twice the average of LMI countries. Population density levels in the other three countries are also well above international averages. 13 Table 1.1: Economic per capita growth by international standards GDP, 2015 GDP per GDP density Population $1.90 Population (PPP, capita, 2015 2015 ($/km2, density poverty rate (million) $billion) (PPP $) thousands) (ppl/km2) (%) Levels Benin 22 2,057 198 11 96 67.8 Burkina Faso 31 1,696 112 18 66 43.7 Cote d'Ivoire 80 3,514 251 23 71 27.9 Togo 11 1,460 196 7 134 49.2 Sub-Saharan Africa 3,718 3,714 157 1,001 42 41.0 Low & Middle Income 61,047 9,911 645 6,159 65 12.6 Annual growth (2010-2015) (percent) (percent) (percent) (percentage point) Benin 4.3 1.5 2.7 3.67 Burkina Faso 5.9 2.8 3.0 -2.32 Cote d'Ivoire 5.8 3.3 2.4 -0.16 Togo 4.8 2.0 2.7 -1.26 Sub-Saharan Africa 4.3 1.5 2.8 -1.56 Low & Middle Income 5.2 3.9 1.3 -1.93 Source: World Bank, 2017. Interestingly, the sub-region’s relatively high population density and high GDP density (defined as GDP per square kilometer), which are often considered favorable element for economic development (World Bank, 2009), do not translate into higher levels of prosperity. In three out of four countries— Benin, Burkina Faso, and Togo—about half of the population lives in extreme poverty, a rate that is even higher than African averages. Even in Côte d’Ivoire, the only middle-income country in the group, with a per capita GDP above US$3,500 PPP per year, one in every four persons still lives on less than US$700 annually. A possible explanation for this mismatch between population density and economic prosperity may be the vast unevenness in economic development within a country’s border. Chapter 3 will explore this aspect further. Data As discussed above, national-level comparisons mask large disparities at subnational levels. This section describes the data used to explore within-country disparities in welfare, poverty, and other development indicators. Statistical Data This study makes use of four recently collected household consumption surveys: the Benin Integrated Modular Household Well-Being Survey (Enquête Modulaire Intégrée sur les Conditions de Vie des Ménages, EMICOV 2015), the Burkina Faso Continuous Multi-Sectoral Survey (Enquête Multisectorielle Continue, EMC 2014), the Côte d’Ivoire Household Living Standards Survey (Enquête sur le Niveau de Vie des Ménages, ENV 2015), and the Togo Basic Well-Being Indicator Questionnaire (Questionnaire des Indicateurs de Base du Bien-être, QUIBB 2015). While earlier household consumption surveys for each country are also available (i.e., Benin EMICOV 2010, Burkina Faso GHS 2009, Côte d’Ivoire ENV 2011, and Togo QUIBB 2011), their lack of comparability with more recent surveys limits our capability to observe changes in geographical patterns 14 of welfare and poverty over time.2 Therefore, we focus on a static analysis of economic geography in this sub-region. We also take advantage of a number of harmonized data sets from the Survey-based Harmonized Indicators Program (SHIP) produced by the World Bank, which aim to compile in a consistent format consumption aggregates and other household indicators such as demographics and assets from household budget surveys in the SSA sub-region. However, our four surveys were collected only recently and have not been fully processed in SHIP at the time of the writing. In general, household consumption surveys, including those used in this report, are designed to produce welfare measures and development outcomes at national and, in some cases, at the first sub-national level (e.g. regions). Disaggregating the data into lower administrative unit levels may pose two risks: lack of representativeness, and imprecise estimates.3 On the one hand, households who live in a small geographical area and were interviewed for the surveys may not represent the wider population. On the other, the limited number of households reporting the information of interest leads to higher odds of ending up with missing information (e.g., access to improved toilets), or, when information is available, wide variance (e.g., outliers). To arrive at a balance between data and administrative coherence such that conclusions can be useful from a policy and development perspective, we focus on second sub-national level data (i.e., communes for Benin, provinces for Burkina Faso, departments for Côte d’Ivoire, and prefectures for Togo). It is important to note that even at these sub-national levels, we can limit but not entirely eliminate the two shortcomings discussed above. We rely on the agriculture and land modules contained in these surveys to capture the geography of agriculture from the perspective of households. Unlike administrative data on agricultural production, our approach is more likely to be biased toward smallholder agriculture instead of large commercial farms. While livestock is a major source of income for households in Sahel regions, the data are not available across all four countries in the sub-region, and this dimension is therefore excluded from our analysis. Geographic Data To construct a market access index, we used the road network provided by DeLorme (2015). While an ideal index should capture access to all modes of transportation (e.g., air, coast, rail, etc.), we did not have access to such data at the time of writing. Thus, our index is limited to reflecting domestic market access to roads. For our multivariate regressions in Chapter 5, we constructed six continuous agro-ecological variables at the administrative unit level for each country:4 temperature, precipitation, soil quality, latitude, elevation, and ruggedness. The temperature variable is a long-run (1960–1990) annual average taken from Hijmans et al. (2005), and precipitation is taken from HarvestChoice/International Food Policy Research Institute (IFPRI) and University of Minnesota (2016), which measures annual average over the period 1960–2014. Soil quality is measured as organic carbon soil content (fine earth fraction) at 60–100cm depth taken from HarvestChoice/IFPRI and University of Minnesota (2016). Ruggedness is based on Nunn and Puga (2012). Elevation, given in meters, is taken from Isciences (2008). Except for HarvestChoice/IFPRI and University of Minnesota (2016), these data consist of a 30-arc-second grid solution, equivalent to 1 x 1 km, thus making it possible to aggregate the data to the administrative unit level. For its part, grid resolution for HarvestChoice/IFPRI and University of Minnesota (2016) data is 5- arc-minute, or roughly 10 x 10 km. Thus, some smaller administrative units will not have data. For such 2 Household consumption surveys are considered comparable if all three of the following criteria are consistent across surveys: (i) the sample size is nationally representative; (ii) the data were collected during the same period; and (iii) the surveys rely on the same reporting instrument and reporting period (Beegle et al., 2016). 3 As a rule of thumb, estimates are considered sufficiently precise if the relative standard error (measured as standard error divided by the mean) is less than 10 percent. 4 Communes for Benin, provinces for Burkina Faso, departments for Côte d’Ivoire, and prefectures for Togo. 15 areas, we impute a given variable as the average of that variable across its neighboring administrative units (Table 1.2). Table 1.2: Geographical data sources Data Sources Administrative boundaries National statistical services, Global Administrative Area Database (GADM) Agro-ecological zones (Benin, Togo) Food and Agriculture Organization (FAO) of the United Nations, Togo’s Ministry of the Environment and Forestry Resources Climate zones Kottek et al. (2006) Elevation Isciences (2008) Livelihood zones (Burkina Faso, Côte Famine Early Warning Systems Network (FEWSNET), d’Ivoire) AGRHYMET Population density at sub-national National statistical services levels Precipitation HarvestChoice/IFPRI and University of Minnesota (2016) Road network DeLorme (2015) Ruggedness Nunn and Puga (2012) Soil quality HarvestChoice/IFPRI and University of Minnesota (2016) Temperature Hijmans et al. (2005) 16 Chapter 2: Geography of Welfare – Three Building Blocks This chapter defines three spatial scales as stepping stones for the spatial analysis of welfare and agricultural activity to follow in Chapters 3 and 4. These scales are based on the three building blocks found in the economic geography literature. We start with the building block found in the traditional economic geography literature: natural endowment. A region is expected to be better off than others if it is endowed with a more favorable agro-ecosystem, more natural resources, or simply a better location. We then add two main elements from any NEG model as well as from any modern theory of location:5 agglomeration economies (Marshall, 1920; Krugman, 1991; Porter, 1998; Henderson, 2014), and access to markets, that is, markets for goods, labor, and ideas (Smith, 1776; Fujita and Thisse, 2002). Finally, we assess what these three elements look like in each country: Benin, Burkina Faso, Côte d’Ivoire, and Togo. The core idea of NEG is that a location is not an isolated geographical area but is affected by its relationships or connections with neighboring locations. Agglomeration economies ensure that economic activity is concentrated in areas that are better located to benefit from increasing returns to scale. Access to markets then captures the levels of transportation costs and the degree of labor mobility between locations. High levels of market access (i.e., free movement of goods and people across space) combined with increasing returns to scale will create spatial disparities in economic activities, and therefore poverty. In this study, we do not cover tangible costs such as road tolls or legal requirements for residency or non-tangible costs such as discrimination or ethnic or religious differences that may be associated with market access. Our main findings are: (a) In all four countries, there seems to exist a tale of two regions—North vs. South for the coastal countries (Benin, Côte d’Ivoire, and Togo), and Center vs. the Rest for landlocked Burkina Faso. (b) Within a country, the North generally has the least favorable agro-ecological characteristics for agricultural activities, while the South is the most endowed. (c) Similarly, agglomeration economies cluster mainly in the South. The exception is landlocked Burkina Faso, where the densest population is in the central region, home of its capital, Ouagadougou. (d) Market access follows the same pattern. Most land areas in the North have very limited access to markets, while high levels of market access concentrate around the capitals, the economic capitals, and along the coast in the South. Natural Endowment In this section, we examine each country’s natural endowments (i.e., agro-ecological endowment), standardize existing classifications of agro-ecological zones, and re-group these into four broad zones ranging from least favorable (Zone 1) to most favorable (Zone 4). This re-categorization allows us to overlay patterns of welfare, poverty, agricultural productivity, and economic activity over agro-ecological zones in a consistent and systematic manner across countries. Agriculture plays a key role in the four countries of interest. This sector generates about a third of GDP value each year. In Togo, nearly half of GDP in 2015 came from agriculture alone. In addition, the agriculture sector provides jobs to about half of the workforce (World Bank, 2017). In an agricultural economy, agro-ecological endowments determine not only what crops are planted or livestock is raised in 5 For a detailed discussion of NEG, see, for example, Fujita, Krugman, and Venables (1999), Fujita and Thisse (2002), Baldwin et al., (2003), Brakman, Garretsen, and Van Marrewijk (2009), Combes, Mayer, and Thisse (2008). For major NEG models, see Krugman (2001), Krugman and Venables (1995), Venables (1996), and Puga (1999). 17 each location but also what returns can be obtained for any crop harvested or livestock herded. These consequently affect the region’s agricultural productivity and welfare (see, for example, Diamond, 1997). Compared to the rest of SSA, West Africa is relatively abundant in precipitation, with most of its land in the savannah and grassland areas, and it benefits from a tropical climate (Map 1.1, Map 2.1, and Appendix A). Within the sub-region, two distinct groups emerge. Being the northernmost and the only landlocked country, Burkina Faso has noticeably different agro-ecological characteristics, being generally drier and less fertile. The rest—Côte d’Ivoire, Benin, and Togo—are coastal and located at the same latitude, thus sharing similar climates and precipitation levels. Map 2.1: Abundant in terms of precipitation levels Source: Funk et al., 2015. A clear picture immediately comes to the forefront: In all four countries, Zone 1 is in the north, while Zone 4 is concentrated in the south. Interestingly, Zones 4 in Benin, Côte d’Ivoire, and Togo are also coastal, which is generally considered an advantage for economic development (Map 2.2). Overviews of each zone are detailed below. 18 Map 2.2: Agro-ecological zones, by country Sources: AGRHYMET, 2016; Dixon and Holt, 2010; FAO, 2001, 2009a, 2009b; Ministère de l’Environnement et des Ressources Forestières, 2003, 2014; Vissoh et al., 2004. Zone 1 represents the Sahelian and Sudanian savannah areas, the driest of all four zones within a country. There is only one rainy season, which is also relatively short. However, precipitation levels vary considerably across countries, starting from the lowest in Burkina Faso to the highest in Côte d’Ivoire. • Burkina Faso: Zone 1 is typical Sahelian, with four months of rainfall per year, accumulating approximately 400–500 mm. The soil is sandy and of poor quality. • Côte d’Ivoire: While the annual precipitation level of 1,000–1,100 mm for five months of the year is lower than those in other zones within Côte d’Ivoire, it is the highest when compared with Zones 1 in Burkina Faso, Benin, and Togo. The area is characterized by savannah, with a mixture of woodland and grassland. • Togo: This zone has very similar characteristics in terms of savannah landscapes and average amount of rainfall, albeit slightly lower, compared to Zone 1 in Côte d’Ivoire. • Benin: The climate is Sudano-Sahelian with a unimodal rainfall pattern of 700–1,000 mm per year. The area is marked with a vast expanse of arable land in ferrosol soil. Zone 2 is also characterized by unimodal rainfall patterns, albeit with slightly higher precipitation levels and a longer rainy season than in Zone 1. The capital of Burkina Faso, Ouagadougou, is located in this zone. • Burkina Faso: This Sudano-Sahelian zone receives about 600–800 mm of rainfall per year. However, it has poor quality soil and faces serious land erosion problems. 19 • Côte d’Ivoire: Vegetation types are the same as in Zone 1, i.e., woodlands, grassland, and savannah. However, precipitation levels are higher, at 1,100–1,300 mm per year. The zone is typically characterized by flat terrain and ferrosol soil. • Togo: The climate is Sudano-Guinean, with savannah landscapes. • Benin: This zone shares similar characteristics with Togo’s Zone 2. Zone 3 receives more plentiful rainfall than the other two zones. • Burkina Faso: The zone transits into a savannah ecosystem and a Sudanian climate. Precipitation levels are 800–900 mm per year (unimodal rainfall curves) along with good soil quality. The zone is also endowed with large forests and vast areas of animal reserves. • Côte d’Ivoire: The zone covers two distinct ecosystems: a Guinean type in the mountains, and a Sudanian type in the flatlands. Average annual rainfall is 1,250–1,500 mm. • Togo: This zone has a Guinean climate and is largely made up of the Togo Mountains, which can reach nearly 1,000 meters in height at Mount Agou. • Benin: Being a transitional zone, it has no clear distinction between the two rainy seasons. The landscape, however, is similar to Zone 2, which is woody savannah with tropical ferruginous soil. Zone 4 has the most abundant rainfall and the most fertile soil. It houses the capital Lomé in Togo and the economic capitals in Côte d’Ivoire and Benin, Abidjan and Cotonou, respectively. • Burkina Faso: This zone shares a similar ecosystem and climate with Zone 3. However, it receives more rainfall, at about 900–1,100 mm per year. • Côte d’Ivoire: The coastal zone receives up to 1,750 mm of rainfall per year. • Togo: This is a coastal zone with a sub-equatorial climate. However, precipitation levels are lower than in other zones, at about 750–1,000 mm annually. • Benin: This zone has a sub-equatorial climate, with two rainy and two dry seasons. The soil type is mostly ferralitic, including relics of forest. Agglomeration Economies The second building block—agglomeration economies—is a vital engine of innovation and growth and plays a powerful role in explaining within-country inequality. As countries develop over time, economic activity clusters in certain locations to take advantage of both economies of scale and knowledge exchanges (Marshall, 1920; Krugman, 1991; Porter, 1998; Henderson, 2014). Greater economic opportunities in turn attract people who migrate in search of jobs, which consequently boosts population density in one area over another (World Bank, 2009). As discussed in Chapter 1, this sub-region is relatively densely populated by African standards. In terms of urbanization, as reported by the various national statistical services, the share of population residing in urban areas in three countries—Togo, Benin, and Côte d’Ivoire—is relatively high, at over 40 percent of the total population, a rate that is higher than the SSA average. However, growth rates for urbanization are slower, with Burkina Faso having the lowest share of population living in urban areas, at less than 30 percent. Nevertheless, the country is catching up, with impressive urban population growth of nearly 6 percent annually (Figure 2.1). 20 Figure 2.1: Higher share of population living in urban areas but slower urban population growth by African standards Source: World Development Indicators 2017. To go beyond the urban-rural dichotomy seen in the literature and, more importantly, to arrive at a consistent and systematic classification across countries,6 we further distinguish localities into four groups: ultra-remote rural, rural, urban, and ultra-dense urban. Density thresholds used to categorize these groups are based on the Organisation for Economic Cooperation and Development—OECD (1994), Uchida and Nelson (2010), and Buys, Chomitz, and Thomas (2005). Specifically: (a) Ultra-remote rural localities are defined as those having fewer than 50 people per square kilometer. (b) Rural areas are locations with population density of between 50 and 150 people per square kilometer. (c) Urban areas are characterized by having population density of between 150 and 300 people per square kilometer. (d) Ultra-dense urban localities are areas with more than 300 people per square kilometer. As shown in Map 2.3, in three out of four countries, namely Benin, Côte d’Ivoire, and Togo, urban and ultra-dense urban localities are concentrated mainly in the South, coinciding with the most favorable agro- ecological zone (Zone 4) as well as coastal areas. Landlocked Burkina Faso is an exception in that the densest part of the country is located in the central region, around the capital Ougadougou. Compared to other countries in the sub-region, Burkina Faso has the most dispersed population. Apart from ultra-dense Kadiogo Province, home to the capital Ougadougou, the rest of the country is made of ultra- remote rural or rural areas. Based on our classification, the country does not even have urban localities. In contrast, Togo’s population is the densest, with the only ultra-remote rural location being found in the Central region, where the Togo Mountains lie. 6 Each national statistical agency has different definition of “urban locality.” 21 Map 2.3: Concentration of agglomeration economies in the South Source: National statistical agencies. Market Access Based on NEG theory, the benefits of agglomeration economies can be further enhanced with the existence of good access to markets for products, labor, and ideas (Mayer, 2008). A region with better market access will attract more economic activity and labor, leading to an agglomeration advantage over time. With increasing returns to scale, the region can then afford to reinvest in market access and further reinforce its advantage. This can start a virtuous cycle of development, which is good for economic growth and poverty reduction but also makes it difficult for disadvantaged regions to catch up (World Bank, 2009). In this section, we start with a brief glance at the current road network, a core factor in market access. We then proceed to calculate each country’s market access index based on the most commonly used modified model of the “classical” approach.7 As highlighted in NEG literature, access to markets must be considered beyond a country’s border. This element is especially crucial for landlocked countries such as Burkina Faso. For Burkina goods to reach new markets and for Burkina people to receive more products from the outside world, there must be good transport connections to neighboring countries. As shown in Map 2.4, several primary roads connect large cities in Burkina Faso to their neighbors’ coastal ports in Benin, Côte d’Ivoire, and Togo. Among the four countries, landlocked Burkina Faso has a relatively extensive primary and secondary road network that extends to all four agro-ecological zones. In contrast, coastal countries such as Côte d’Ivoire 7 For more details of the classical and modified models, see, for example, Deichmann (1997) and Lall, Shalizi, and Deichmann (2004). 22 and Benin concentrate domestic transportation systems in coastal and dense areas (Zone 4) and neglect remote regions (Zones 1 and 2). Map 2.4: Road networks Source: DeLorme, 2015. To measure market access, we first follow the classical model in the literature, as follows: Domestic market access for a given location along a road network is a function of the weighted sum of populated locations of all other locations discounted by travel time on the road.8 = ∑ − where is market access in location i, is the population in location j, is travel time between locations i and j, and is a trade elasticity parameter. We then apply the most commonly used modified model as it is more relevant to countries with geographically limited data on populated locations.9 − − ( ⁄ ) = ∑ . 22 where is the population in location j, is travel time between locations i and j, and a and b are trade elasticity parameters based on Deichmann (1997). We then summarize market access at an administrative 8 Examples from the literature with similar market access include Harris (1954), Hanson (2005), Emran and Shilpi (2012), Jedwab and Storeygard (2015), Blankespoor et al. (2016), Berg, Blankespoor, and Selod (2016), and Donaldson and Hornbeck (2016). 9 See Lall, Shalizi, and Deichmann (2004), Yoshida et al. (2009), and Ballon et al. (n.d.). 23 level for each country by converting the market access results to an inverse distance weighted grid and taking the mean of the grid in the administrative level. The spatial distribution of the market access indicator is presented in Map 2.5. (Appendix B provides further details of the construction of our market access indicator.) There are two key limitations to our model. First, we only consider land transportation (e.g., road networks), thus potentially underestimating market access index in coastal areas, where sea access is available. Second, our model computes domestic market access with locations across borders not being taken into account. As a result, market access index of areas along a country’s border may also be underestimated. Two striking facts emerge. First, across all four countries, the North generally has limited access to market. Second, areas with high market access cluster in the South, around the capitals and the economic capitals—Cotonou (Benin), Ouagadougou (Burkina Faso), Abidjan (Côte d’Ivoire), and Lomé (Togo). However, not all coastal areas are born equal, as shown by Côte d’Ivoire, where the western coastal side of the country does not enjoy the same levels of market access as the eastern side, at least not until it reaches closer to the border with Liberia. Map 2.5: Concentration of high market access in the South, around capitals and economic capitals (a) Benin (b) Burkina Faso (c) Côte d’Ivoire (d) Togo Source: Authors’ calculations based on data from De Lorme (2015). 24 Chapter 3: Spatial Disparities in Welfare and Poverty This chapter visually presents the geographical distribution of welfare and poverty and relates it to the three key elements of economic geography: natural endowment, agglomeration, and market access. As described in Chapter 2, spatial disparities in welfare could be a natural outcome of the cycle of development: as a country develops, economic activity concentrates in certain regions to take advantage of economies of scale, with these regions in turn attracting more people looking for job opportunities, which consequently increases population density. With increasing returns to scale, these regions can continue to invest in access to markets and thus further reinforce their advantage. However, this development process makes it more difficult for poorer areas to catch up. If economic development is inevitably uneven within a country, where are the leading and lagging areas located, and where do the majority of the poor live? These are among the questions we aim to answer in this chapter. Within the domain of welfare and poverty, we focus on four important dimensions: leading and lagging regions, poverty measures (including poverty rates, poverty mass, and poverty density), access to basic services, and profile of the poor. Arguably, the stylized facts presented in this chapter are useful for policy makers for several reasons. First, they can guide budget allocation across administrative units by identifying which regions have fallen behind in terms of economic development, which regions have forged ahead, and more importantly, the magnitude of the income gap between them. Second and along the same lines, the geographical targeting of programs designed to alleviate poverty can benefit from the identification of geographical areas with high prevalence of poverty, or poverty rates, defined as the share of population living below US$1.90 a day at 2011 PPP. In addition to poverty rates, information on the number of the poor, or poverty mass, is handy when it comes to cost estimates of a social policy targeted to the poor, such as social safety net programs. Third, public investments in service delivery programs designed for the poor can be prioritized accordingly. In this context, services may come in many forms and include social services such as primary education for all, economic services such as irrigation systems for poor farmers, or information services such as mobile phone coverage. The spatial distribution of poverty density, defined as the number of poor people per square kilometer, and maps of current public services coverage are critical for policy makers to decide whether a new service delivery program can be offered or an existing program can be expanded in a cost-effective way. If so, how many locations can the programs reach, and where are these locations to be found? The coverage of such programs depends heavily on the projected costs (e.g., upfront investment such as schools, piping for water connections, electricity lines and poles, etc.), which in turn are largely determined by the density of users and the current status of public service coverage. Finally, by assessing how the characteristics of the poor differ across space from food consumption patterns to household demographics, we aim to help governments quickly identify affected groups in cases of shock (e.g., rise in commodity prices, etc.) or policy reforms related to food products, such as maize subsidies. Within a country, the poor display distinct characteristics and face different challenges in each location. For example, factors with significant impact on the poor’s welfare in the North, such as maize price, may play a lesser role than those in the South. To preview our main findings: (a) There is a large income gap between leading and lagging areas. This gap is highest in Benin and lowest in Burkina Faso. (b) Many of the leading areas have not reached their full potential. In other words, many have not yet maximized the benefits of agglomeration economies (especially in Burkina Faso and Côte d’Ivoire) or of market access. 25 (c) Within a country, there is wide variation in poverty rates, such that the North is markedly poorer than the South (except in Burkina Faso). (d) The poverty mass—the number of poor people—is highest in low-density areas (in Burkina Faso, Côte d’Ivoire, and Togo). This pattern suggests that the cost for service delivery programs to physically reach the poor could be relatively high, especially as access to public services such as improved toilets, piped water, and electricity differs greatly across space, with the North having lower coverage than the South. (e) In each country, there exist geographical pockets of poverty, which may be resistant to policy- induced change. Related to our main findings #3 and #4 above, these areas are characterized by a combination of high poverty rates and low poverty densities. Therefore, the unit cost of a poverty- targeted program could be extremely high in these areas. Given budget constraints, the government may not be able to reach this population group. (f) Looking beyond monetary poverty (i.e. US$1.90 a day at 2011 PPP), the quality of life of the poor measured by the extent of food intake diversification and housing conditions varies across space, with geographical patterns mimicking those observed with monetary poverty. Leading and Lagging Areas As discussed in Chapter 2, the accumulation of wealth in one area but not in another may be a natural outcome of the development path (made plausible through the evolution of agglomeration and expansion in market access). The question is therefore: What regions are benefiting from the fruits of development and which are falling behind? More importantly, how wide is the gap between them? In fact, there are only a few leading areas and many lagging ones (Map 3.1). We define a location as “leading” if per capita consumption for an average household living there is higher than the national average. Out of 107 departments in Côte d’Ivoire,10 only 19 are leading. This figure is 14 out of 77 communes for Benin, 11 out of 45 provinces for Burkina Faso, and 9 out of 36 prefectures for Togo.11 10 Although Côte d’Ivoire has 108 departments, the 2001 ENV survey covers only 107 of them. 11 For Benin, leading communes are Abomey, Abomey-Calavi, Adjarra, Bohicon, Cotonou, Dassa-Zoumé, Houeyogbe, Natitingou, Ouèssé, Parakou, Porto-Novo, Sakété, Savé, and Sèmè-Kpodji. Leading provinces in Burkina Faso are Boulgou, Comoé, Houet, Kadiogo, Nahouri, Noumbiel, Oudalan, Poni, Sanmatenga, Séno, and Yagha. For Côte d’Ivoire, Abidjan is one of leading departments, the rest being Abengourou, Aboisso, Adzopé, Bangolo, Bettié, Blolequin, Bouaflé, Dabou, Duékoué, Gagnoa, Grand-Bassam, Guéyo, Guiglo, San-Pedro, Sikensi, Tabou, Yamoussoukro, and Zuénoula. For Togo, leading prefectures are Bassar, Cinkassé, Danyi, Golfe, Lacs, Lomé, Ogou, Tchaoudjo, and Vo. 26 Map 3.1: Cluster of leading areas in the South, around the capitals and the economic capitals, or along country borders Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. Note: A location is defined as “leading” if consumption per capita for an average household living there is higher than the national average. The limited number of areas with per capita consumption above the national average suggests a large wealth gap between leading and lagging locations. As shown inFigure 3.1, the difference in income between the top three leading areas and the bottom three lagging ones could be as high as a factor of 7 (in the case of Benin). In this regard, Burkina Faso is the least unequal country, with a ratio of approximately 3.5 between the province with the highest income and the poorest one. Figure 3.1: Large wealth gap between leading and lagging areas 27 Sources: Authors’ calculation based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. Surprisingly, many of these leading areas have not yet maximized the benefits of agglomeration economies. This observation is especially clear in Burkina Faso and Côte d’Ivoire (Figure 3.2). In a country with sparse population such as Burkina Faso, a majority of the better-off provinces are still located in ultra- remote areas. The fact that these provinces have not taken advantage of increasing returns to scale by urbanizing could partially explain why the wealth gap between rich and poor provinces is relatively low in Burkina Faso. We notice a similar pattern in Côte d’Ivoire, where nearly half of the leading departments are located in rural areas. However, some of the leading departments in the country host large-scale farmers, who need land of considerable size for their operations, thus explaining their low-density locations. It is important to point out a mixed picture in Benin. While some leading communes have low population density, many lagging ones are situated in either urban or ultra-dense urban areas. Why, therefore, do these locations, which could enjoy the benefits of agglomeration externalities, remain poor? In fact, these dense but lagging communes cluster in the South near the coast of Benin (Map 2.3 and Map 3.1). While the poor might enjoy agglomeration economies, the nature of the economic activities in this area, which consists mostly of informal trading with Nigeria, might attract large populations of poor migrants (Golub, 2012). Figure 3.2: Many leading areas in low-density locations 28 Sources: Authors’ calculation based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. Similarly, we find a mixed pattern between leading and lagging areas and market access. Although a region with higher per capita consumption is often shown to have better market access, this pattern does not always hold. Figure 3.3 illustrates how many leading administrative units in fact have low market access (defined as having market access below the average value across all four countries) and vice versa. As mentioned in Chapter 2, a limitation of our market access index results from underestimates of market access values in administrative units along a country’s border and along the coast. This could explain why some leading administrative units do not have high market access. Figure 3.3: Low market access in many leading areas Sources: Authors’ calculation based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. 29 Poverty Rates, Poverty Mass, and Poverty Density The previous section discussed leading and lagging areas in terms of income.12 We now turn to an equally pressing concern: Which areas experience widespread poverty incidence and which ones do not? Here, we use the US$1.90-a-day poverty line at 2011 PPP to define poverty. Here again, we observe a tale of two regions: North vs. South (Map 3.2). Within each of the three coastal countries—Benin, Côte d’Ivoire, and Togo—the North, corresponding to the two least favorable agro- ecological zones (Zone 1 and 2) is markedly poorer than the South, which comprises the two most favorable zones (Zone 3 and 4). In Togo particularly, poverty in the far North may be more than three times as high as in the far South. However, the pattern of poverty is reversed in Burkina Faso. One possible explanation is the possession of livestock among Burkina inhabitants in the North, which is the case in neighboring countries such as Mali and Niger. Map 3.2: In three out of four countries, the North is remarkably poorer than the South. Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. Not surprisingly, leading areas have the lowest poverty rates and are clustered in the South, around the capitals and the economic capitals, and along the country’s border (Map 3.3). This pattern of poverty is consistent with the geographical distribution of leading and lagging regions described previously. However, what is alarming is the large variation in poverty incidence between leading and lagging areas. In Benin, poverty rates can vary between 20 percent in Cotonou and nearly 100 percent in the three most lagging communes (Cobli, Copargo, and Boukoumbé). A similar range is observed in Togo, or between 15 percent in the top three prefectures (Golfe, Lacs, and Lomé) and above 90 percent in the bottom three (Tandjoaré, Akebou, and Doufelgou). For Burkina Faso, the figure ranges from about 10 percent in 12 Measured as per capita consumption. 30 the richest provinces (Noumbiel, Nahouri, and Kadiogo) to above 80 percent in the most disadvantaged ones (Komandjoan, Zondama, and Sourou). Côte d’Ivoire tells the same story, with poverty rates between 8 percent in the top three departments (Guéyo, Abidjan, and Tabou) and over 80 percent in the bottom three (Tengrela, Sipilou, and Oumé). Figure C. in Appendix C provides details of poverty rates at sub-national levels for each country. Map 3.3: Lower poverty rates in areas around the capitals and the economic capitals, and along the country border Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. A relevant consideration for policy makers is not only the issue of where poverty rates are high but also where the poverty mass is located. Here, we define poverty mass as the number of poor people. A location with a lower poverty rate does not necessarily imply that it has fewer poor people when population is taken into account. In fact, relatively better-off locations, such as the capitals or the economic capitals of countries and areas in the South, have high poverty mass. In Benin and Côte d’Ivoire, nearly half of the poor congregate in Zone 4 (the South). Similarly, in Burkina Faso, about half of the poor population lives in Zone 2 in and around the capital. Meanwhile, Zone 3 in Togo is home to about 40 percent of the country’s poor. At sub-national levels, some leading areas in Benin and Côte d’Ivoire host the highest number of poor people.13 These are Abomey-Calavi commune, suburban Cotonou in Benin, and Abidjan department in Côte d’Ivoire. In contrast, in Burkina Faso and Togo, the largest poor population is concentrated in two of the most disadvantaged areas in terms of income: Yatenga province in Burkina Faso, and Oti prefecture in Togo. 13 Compared to other administrative units in the same country. 31 Nevertheless, most of the poor still reside in either ultra-remote rural or rural areas, where density is lower than 150 people per kilometer (Figure 3.4).14 This strong pattern seen in Burkina Faso, Côte d’Ivoire, and Togo, where at least three out of four poor people live in low-density locations, suggests that the cost of physically reaching the poor could be relatively high in these three countries. Figure 3.4: Majority of the poor living in low-density areas Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. The final—though no less important—dimension of poverty discussed in this section is poverty density. Similar to population density, poverty density is defined as the number of poor people per square kilometer (Map 3.4). Not surprisingly, the highest density of the poor is found in the capitals or the economic capitals. Across countries, Cotonou commune in Benin is the densest, at nearly 1,700 poor people per square kilometer, Lomé commune in Togo follows, with poverty density of 1,500, and Abidjan department in Côte d’Ivoire houses about 200 poor people per square kilometer. In sparsely populated Burkina Faso, where poverty density is relatively low, even the densest province—Zondoma—has only 78 poor people per square kilometer. 14 Following our agglomeration classification in Chapter 2. 32 Map 3.4: High poverty density in and around the capitals and the economic capitals, and in the South Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. It is important to note stark variation in poverty density across administrative units, which affects the cost of various government programs delivered to each location. In Benin, the difference between Cotonou commune, with the highest number of poor people per square kilometer, and Karimama commune, with the lowest poverty density, is strikingly high, or nearly 200 times. This ratio is 120 for Togo, 73 for Côte d’Ivoire, and 31 for Burkina Faso. Annex C.2 lists poverty density for each administrative unit in the four countries of interest. Access to Services Following our discussion of how density of users could be one of the decisive factors affecting service delivery coverage, we now examine disparities in the provision of services across space. In this section, we focus on four service delivery programs available in our data sets, namely access to electricity, piped water, improved toilet facilities, and mobile phone coverage.15 We note that the share of the population having access to electricity, piped water, and improved toilets increases with population density, with the highest coverage observed in ultra-dense urban locations while the lowest is seen in ultra-remote rural areas (Figure 3.5). This pattern is consistent with recent research that shows a close correlation between access to basic services and population density (Gollin, Kirchberger, and Lagakos 2016). Across agglomerations and countries, mobile phone coverage is 15 Improved toilet facilities include flush toilets, ventilated improved pit latrines, composting latrines, and pit latrines. 33 homogenously high by developing country standards: on average, at least 7 out of every 10 people in the sub-region have a mobile phone. This uniformly wide coverage could be explained by the fact that mobile phone networks require relatively low costs to reach end-users compared to other service delivery programs such as electricity or water. Figure 3.5: Large gaps in public service coverage between the most sparsely and most densely populated areas Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. However, there are wide differences in access to services (piped water, improved toilets, and electricity) across countries. Côte d’Ivoire, with a national income nearly twice that of each of the other three countries, takes the lead, with Burkina Faso trailing far behind. In Côte d’Ivoire, one third of the population has access to piped water, two thirds use electricity in their house, and three quarters have access to an improved toilet facility. For Burkina Faso, these figures are 11 percent, 47 percent, and 15 percent, respectively. Within country, we observe a similar tale of two regions, as discussed in the previous sections. While nationwide mobile phone coverage is more or less equally distributed, coverage of piped water, improved toilets, and electricity appears to be divided into two regions (Map 3.5). In Benin, Côte d’Ivoire, and Togo, the dividing line for the coverage of piped water, improved toilets, and electricity lies between the North (Zones 1 and 2) and the South (Zones 3 and 4), with a larger share of the population in the South having access to public services. In Burkina Faso, the division is between the Center South (Zones 2 and 4) and the rest of the country). 34 Map 3.5: High variation in access to public services across space (except cellphone coverage) Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. At the sub-national level, we document a wide geographical disparity in access to services, which goes hand in hand with spatial differences in population density and poverty rates (Map 3.3 and Map 3.6). Understandably, areas characterized by a combination of high poverty incidence and low population density tend to have low services coverage (except for mobile phone networks). The variation in services coverage can be large across space. In Benin, for example, merely 2 percent of the population in Karimama commune, one of the most sparsely populated and poorest areas, have access to electricity and improved toilets. Meanwhile, the economic capital, Cotonou, is at the other end of the spectrum, with more than 80 percent of residents having such access. A similar story is observed in Côte d’Ivoire, where one of the poorest and most sparsely populated areas, Sipilou department, shows less than 15 percent of its population having electricity, piped water, and improved toilets compared to almost 100 percent of the population in the economic capital, Abidjan. Togo shares the same pattern, with nearly no one in Blitta prefecture having electricity or an improved toilet at home, while over 90 percent of the residents of the capital, Lomé, do. In Burkina Faso, Loroum province provides electricity and piped water to almost none of its inhabitants, while this figure is above 50 percent for people living in Kadiogo province, where the capital, Ougadougou, is located. 35 Map 3.6: Low service coverage in areas with high poverty incidence Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. Quality of Life and Characteristics of Poor People Given that the previous sections show evidence of vast spatial disparities in poverty and access to services, we would expect the quality of life for poor people to vary greatly as well. While people may be considered poor based on monetary standards (i.e. living below US$1.90 a day at 2011 PPP), their living standards and demographic characteristics may differ across space and reflect conditions in the location where they live. To examine the poor’s quality of life, we explore the spatial differences in two important elements of daily life: the composition of the food basket, and housing conditions, as the poor spend a large share of their income on food, and their house (and land) may be their most valuable asset. In this section, we focus on housing because it is particularly relevant for poor people living in urban areas. Finally, we provide a snapshot of poor households’ demographics. Our first measure of quality of life is the diversification of food intake. Keeping other factors constant, it is arguable that individuals are better off when they have a wider range of options for food items to be consumed, which enriches their diet and taste. We use the Herfindahl Index (HI), also known as the 36 Hirschman Index or Hirschman-Herfindahl Index as an inverse measure of variety in food consumption.16 The HI ranges from 1/n to 1, and reaches a maximum value of 1 if the share of consumption is entirely concentrated on a single food item. In other words, the HI measures diversity, where the higher the value of the index, the lower the diversity (Lee and Brown, 1989). Map 3.7a suggests that in all countries, people in the South enjoy a more diverse food basket than their fellow citizens living in the rest of the country, with a lighter color indicating a lower HI.17 Map 3.7b provides a complementary story: people in the South tend to consume less from their own production (from a limited number of crop products). As Chapter 2 pointed out, Southern areas generally have better market access and higher population densities. Thus, their inhabitants rely less on subsistence food production and more on tradable food products. These observations are consistent with the literature on taste for variety: people have more choices (of food intake, jobs, and so on) when they have better access to market or live in agglomerations (Krugman, 1996; Fujita, Krugman, and Venables, 1999). Map 3.7: Higher diversity in food consumption basket and lower food share from own production in the South (a) (b) Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. Next, we examine the types of food that make up the highest share of the poor’s food consumption. We first calculate the share of the food budget for each item for each household in order to identify the most commonly consumed food products. We then estimate the share of the poor population consuming these top-ranked food items. Map 3.8 presents our results. In terms of products, the poor consume similar food items throughout Benin, even though the four agro- ecological zones reveal different shares of the poor population consuming each product. For example, a larger share of the poor have corn in their diet in Zone 4 than in the rest of the country. However, corn is the top-ranked food item for poor people overall. In Zones 1, 2, and 3, the next-ranked food commodity is yam, while it is local rice for those living in coastal area (Zone 4). Burkina Faso exhibits a distinct pattern, 16 The HI is calculated as the sum of squared food shares: ℎ = ∑ 2 =1 ℎ , where the HI of household h is the sum of the budget shares s of each individual food item i consumed in household h. The HI ranges from 1/n to 1. 17 We also note that food diversity measured by the HI should not be compared across Benin, Burkina Faso, Côte d’Ivoire, and Togo because the number and types of food commodities examined in household consumption surveys are not the same across countries. 37 with food consumption for most poor people in the North (Zones 1 and 2) consisting of sorghum and millet, while the food items for those in the South (Zones 3 and 4) include corn and rice. In Côte d’Ivoire, most of the poor in Zone 1 consume yam, while the poor in the rest of country eat local rice. Those living in the coastal South (Zone 4) also have imported rice in their food diet. Interestingly, in Togo, poor inhabitants of the northernmost areas (Zone 1) consume corn and local rice. Meanwhile, those living in the other three zones add dried fish and imported rice to their diet. Moreover, the share of the poor population stating dried fish as their most commonly consumed food product increases as we get closer to the coastal areas. Map 3.8: Variation in key food consumed across space Benin Burkina Faso Côte d’Ivoire Togo Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. 38 Not only do the poor in the South enjoy a more diverse food intake, they also have better housing conditions. Table 3.1 illustrates how the share of the poor population having concrete or tile roofs or cement or brick walls increases significantly from the North (Zone 1) to the South (Zone 4). In all countries, almost no one in Zone 1 lives in a house with either concrete walls or tile roof. However, this share of the poor population in Zone 4 can be as high as 14 percent, as in the case of Togo. Similarly, the difference between Zone 1 and Zone 4 in terms of share of the poor population with cement or brick walls can be from 2 times in Benin to nearly 10 times in Togo. We observe the same pattern across agglomeration types, where poor people living in ultra-dense urban areas have significantly more chances to live in a house with either concrete or tile roof or cement or brick walls. However, poor people in the South and in ultra-dense urban areas are less likely to own their house. One plausible reason could be the relatively high real estate prices in urban areas, such that for the same levels of income, a poor person in the city cannot afford to buy or build a house compared to a similarly poor person living in a rural area. Another possible explanation could be that poor population in urban areas may comprise migrant workers who need temporary housing. Table 3.1: Better housing conditions for poor households in urban areas or in favorable agro-ecological zones Agro-ecological zones Agglomeration types Ultra- Ultra- Zone 1 Zone 2 Zone 3 Zone 4 remote Rural Urban dense rural urban House ownership Benin 0.939 0.952 0.874 0.880 0.924 0.929 0.942 0.842 Burkina Faso n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. Côte d’Ivoire 0.608 0.784 0.690 0.586 0.777 0.667 0.633 0.358 Togo 0.496 0.626 0.571 0.238 0.715 0.517 0.371 0.223 Concrete roof Benin 0.006 0.008 0.010 0.028 0.006 0.007 0.025 0.032 Burkina Faso 0.000 0.003 0.003 0.002 0.001 0.003 n.a 0.015 Côte d’Ivoire 0.011 0.002 0.007 0.007 0.008 0.004 0.011 0.014 Togo 0.003 0.004 0.001 0.138 0.000 0.004 0.050 0.187 Concrete walls Benin 0.19 0.40 0.48 0.45 0.33 0.42 0.26 0.55 Burkina Faso 0.02 0.05 0.06 0.12 0.05 0.08 n.a. 0.16 Côte d’Ivoire 0.59 0.54 0.39 0.50 0.41 0.46 0.56 0.78 Togo 0.06 0.19 0.21 0.62 0.09 0.16 0.28 0.80 Source: Authors' calculations using Benin EMICOV 2015, Burkina Faso E MC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. Lastly, we consider how poor households’ demographics vary across space. In these predominantly agricultural economies, the number of able-bodied household members plays a critical role in determining the scope of households’ economic activity and, consequently, their poverty status. As expected, in areas where natural circumstances are harsh and labor productivity is low, households need more members to generate income. In fact, a typical poor household in landlocked Burkina Faso, on average, has close to 10 members, nearly twice as many as the number in their coastal neighbors, Benin, Côte d’Ivoire, and Togo. Within country, poor households in the most disadvantaged agro-ecological zone (Zone 1) or in ultra- remote rural areas tend to have larger household size and higher share of young dependents than their counterparts living in more favorable zones (Zone 2, 3, and 4) or in denser-population areas. 39 Table 3.2: Fewer family members, lower dependency rates, more likely to be female-headed households, and less likely to have no education among poor households in urban areas and favorable zones Agro-ecological zones Agglomeration types Ultra- Ultra- Zone 1 Zone 2 Zone 3 Zone 4 remote Rural Urban dense rural urban Poor households Household size Benin 6.0 6.1 5.4 4.7 6.05 5.50 4.69 4.70 Burkina Faso 10.0 10.7 9.1 8.7 9.60 10.13 n.a. 8.19 Côte d’Ivoire 5.6 6.2 4.4 5.2 5.47 5.00 4.79 5.96 Togo 6.9 5.9 5.6 5.1 6.46 6.05 5.51 5.33 Number of working-age members Benin 2.45 2.48 2.41 2.14 2.52 2.35 2.15 2.18 Burkina Faso 4.05 4.49 3.54 3.81 4.01 4.24 3.91 Côte d’Ivoire 2.42 2.78 2.15 2.51 2.43 2.33 2.26 3.09 Togo 3.00 2.64 2.55 2.43 2.86 2.66 2.50 2.72 Share of household members under 15 Benin 0.53 0.52 0.48 0.47 0.51 0.50 0.47 0.47 Burkina Faso 0.53 0.51 0.56 0.51 0.53 0.51 n.a. 0.46 Côte d’Ivoire 0.50 0.49 0.43 0.45 0.48 0.45 0.45 0.46 Togo 0.51 0.46 0.46 0.40 0.48 0.48 0.42 0.40 Head of poor households Male Benin 0.92 0.90 0.80 0.76 0.89 0.83 0.78 0.75 Burkina Faso 0.92 0.89 0.90 0.86 0.90 0.87 n.a. 0.86 Côte d’Ivoire 0.69 0.87 0.83 0.82 0.81 0.83 0.79 0.76 Togo 0.85 0.69 0.68 0.64 0.74 0.72 0.68 0.76 Married Benin 0.89 0.86 0.82 0.82 0.87 0.84 0.82 0.82 Burkina Faso 0.94 0.88 0.93 0.87 0.88 0.90 n.a. 0.74 Côte d’Ivoire 0.62 0.89 0.64 0.71 0.77 0.69 0.71 0.64 Togo 0.81 0.71 0.82 0.69 0.83 0.77 0.75 0.74 No education Benin 0.93 0.86 0.76 0.63 0.88 0.81 0.69 0.61 Burkina Faso 0.95 0.92 0.93 0.88 0.90 0.93 n.a. 0.87 Côte d’Ivoire 0.75 0.88 0.62 0.68 0.81 0.68 0.71 0.56 Togo 0.66 0.45 0.40 0.33 0.36 0.54 0.45 0.19 Source: Authors' calculations using Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. When we look at the demographic characteristics of heads of poor households, a pattern emerges: very few poor households are female-headed in ultra-remote rural areas or in Zone 1. The share of poor households headed by a female increases for those living in better-off locations, such as Zone 4 or ultra-dense areas. This observation echoes results of recent studies of female-headed households, which 40 found that in poor areas characterized by a scarcity of jobs or with economic activity heavily dependent on agriculture, widowed or divorced women cannot afford to be on their own and therefore tend to merge with other households (e.g., sons’, brothers’, or through remarriage, etc.). Meanwhile, in relatively richer areas, where more employment opportunities are available, we observe more female-headed households (van de Walle, 2013; 2015). This pattern is consistent with the fact that the share of married heads of poor households is lower in resource-endowed or dense areas. It is interesting to note a smaller share of poor household heads with no education in more favorable locations such as Zone 4 or ultra-dense cities. This suggests two conclusions: poor people with some education may migrate to better-off areas, or there may not be enough jobs to absorb the poor labor force with some education in cities. This problem is more pronounced in Togo, where only 19 percent of poor households in ultra-dense cities have no education. This means that although most of them have some education, they are still poor. 41 Chapter 4: Geographical Differences in Agricultural Activity In this sub-region as around the world, agriculture is a cornerstone of the economy. The agriculture sector forms a large part of the economy and retains great potential to foster inclusive growth. In fact, this crucial link between improvements to the agriculture sector and reductions in poverty is often observed. In 2015, the agriculture sector accounted for roughly a third of GDP for the countries in this sub-region, or 25, 34, 20, and 41 percent for Benin, Burkina Faso, Côte d’Ivoire, and Togo, respectively. The size of the agriculture sector in these countries is comparable to that in low-income countries, where 31 percent of GDP is attributed to agriculture, and it is even larger when compared to Sub-Saharan Africa (SSA), where only 18 percent of GDP is attributed to agriculture, and especially when compared to the rest of the world, where only 5 percent of GDP is attributed to agriculture. The agriculture sector also provides incomes and employment to the vast majority of the population. For example, in Côte d’Ivoire, agriculture employs 67 percent of households, and subsistence farming employs 85 percent of the population. In Burkina Faso, agriculture is the main economic activity for 70 percent of households, employing around 80 percent of the working population. Similarly, in Benin, agriculture accounts for 70 percent of employment. Given the size of the agriculture sector as a share of GDP and employment, it follows that the agriculture sector can—and should—play a potentially important role in growth and poverty reduction. Figure 4.1: Percentage of population engaged in agriculture, by poor and non-poor 100 94.4 % of individuals in agriculture 78.7 80 78.6 77.9 67.5 66.5 62.1 60.3 60 58.2 55.4 51.0 51.4 44.7 43.0 41.1 40 20 0 CIV BFA TGO BEN Sub-Region All Non-Poor Poor Source: Authors' calculations using Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. Beyond its implications for food security, a focus on agriculture and agricultural productivity is crucial to poverty reduction for other reasons. First, it makes sense to improve the productivity of a sector on which most of the poor rely. Among the set of countries in this sub-region, compared to the non-poor, more of the poor are engaged in agriculture. In Togo, for example, among the non-poor, around 40 percent engage in agriculture, while among the poor, close to 80 percent do so. Prior work suggests that agricultural growth reduces poverty by an amount three times greater than growth in other sectors (Christiaensen, Demery, and Kuhl, 2011). Second, improving agricultural productivity can foster structural transformation and manage the urban transition by increasing incomes and promoting non-farm jobs and enables people to move out of agriculture over time (Gollin, Lagakos, and Waugh, 2014; McMillan, 2014). 42 This chapter explores the geography of poverty in this sub-region by examining the geography of agriculture. It focuses on the relationship between agro-ecological zones and key aspects of agriculture, including productivity in crop production, use of agricultural inputs and land tenure security, and sale of output. The agricultural and land modules contained in the most recent household surveys are used for each of the four countries, thus giving an opportunity to make further connections between agricultural activity and poverty at the household level. A key distinguishing factor in our analysis of agricultural activity is precisely that we rely on the agriculture and land modules contained in household surveys as opposed to administrative data on agricultural production. Thus, although this is unlikely to adequately capture the agriculture sector of a country, it is more likely to capture the sector from the perspective of households and is therefore more likely to be biased toward smallholder agriculture as opposed to large-scale commercial farming. However, we should warn that despite the sector’s importance in this sub-region, as, for example, in northern Burkina Faso, livestock and transhumance farming is beyond the scope of this study. Our findings should therefore be interpreted with this limitation in mind. More broadly, mapping productivity, input use, and output sales across agro-ecological zones will enable policy makers to better target agricultural interventions aimed at increasing growth and reducing poverty. We summarize the key findings in this chapter as follows: (a) Relative to the service and industrial sectors, the agriculture sector employs a larger share of individuals in all zones, except in the coastal Zone 4 in Côte d’Ivoire and Togo. The dominance of agriculture is most pronounced in Burkina Faso, where the share of employment in agriculture is 95 and 92 percent in Zones 1 and 3, respectively. Across zones, employment in industry is lowest (under 10 percent in many zones), even though variation in employment in industry is lower in Côte d’Ivoire, suggesting that there is a more equal geographic spread of opportunities for employment in industry there than elsewhere in the sub-region. (b) There is a link between poverty and involvement in agriculture as well as between expenditure and agricultural productivity. This suggests that the geography of poverty is intertwined with the geography of agriculture. Compared to the non-poor, a larger proportion of the poor are engaged in agriculture. In Benin, for example, while 40 percent of the non-poor engage in agriculture, 80 percent of the poor do so. In Burkina Faso, about 95 percent of the poor engage in agriculture. However, among those in agriculture, those with higher maize yields show higher per capita expenditure. Thus, although there is a high proportion of poor in agriculture, improving farm yields can improve their conditions. (c) The agriculture sector varies across countries, with a much more developed sector in Côte d’Ivoire and a much weaker sector in Burkina Faso. In Côte d’Ivoire, the crops that are most prevalent (i.e. planted by most individuals) are a combination of cash crops (crops produced for commercial value) such as cacao and cashew, and food crops such as yam and maize, while food crops are dominant in Burkina Faso, Togo, and Benin. Comparing Burkina Faso to Togo, there is more market exchange of food crops in Togo. For example, while 10 percent of sorghum growers in Burkina Faso sell some of their crop, 57 percent of sorghum growers in Togo do so. Moreover, the zone with the lowest maize yield in Côte d’Ivoire has higher yields than the zo ne with the highest maize yield in each of the other countries. Although cotton production exists across all countries, its intensity varies within countries. In agro-ecological zones, where cotton is not produced to any great extent, other cash crops come to the fore, except in Burkina Faso. (d) Within countries, agricultural performance varies across zones in the sense of higher yields or higher revenues. Better agro-ecologically endowed zones do not necessarily outperform other zones. Zones with a key cash crop, particularly cotton, tend to have higher revenues. In Côte d’Ivoire, Zone 1 is largely disadvantaged, with low yields and low revenues, while Zone 2, 43 which produces cotton, has the highest revenues from crop sales. Similarly, in Burkina Faso, Zone 4, which produces cotton but is also the better agro-ecologically endowed zone, has the highest revenues from crop sales, while Zone 1 has the lowest revenues from sales but better maize yields than Zones 2 and 3. In Togo, Zone 2, which spends a great deal on inputs and also has higher maize yields, has lower revenues from crop sales, while Zone 1 has the highest revenues from sales. In Benin, cotton yields are highest in Zone 1, while maize yields are highest in Zone 2. In Appendix D, a table is provided which provides a brief summary of each zone in each country. Employment in Agriculture Across all countries, the agriculture sector employs a larger share of the population in all agro- ecological zones, except in Zone 4 of Côte d’Ivoire and Togo, where the services sector employs a larger share. The coastal Zone 4 of Benin may be less similar to Zone 4 in Côte d’Ivoire and Togo at least partly because Zone 4 in Benin presents much greater diversity as it consolidates information on three different livelihood zones. Note, however, that while agriculture still employs a larger share in Zone 4 in Benin, the industrial and services sectors combined employ more than the agriculture sector. Map 4.1: Employment in agriculture relative to other sectors Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. Across countries, Zones 1 and 3 in Burkina Faso have distinguishably higher shares of employment in agriculture, while Zone 4 in Togo has a distinguishably lower share. In Burkina Faso, the share of employment in agriculture is 95 and 92 percent in Zones 1 and 3, respectively. In Benin, the share of employment in agriculture is lowest in Zone 4 (39 percent), whereas it is highest in Zone 2 (77 percent). In Togo, there is a noticeably lower share of employment in agriculture in Zone 2, where only 38 percent are employed in agriculture. Additionally, in Togo, only 10 percent are employed in agriculture in Zone 4, the lowest share of agriculture in employment across all zones in all four countries. Similarly, Côte d’Ivoire shows large shares of employment in agriculture in Zone 1 (54 percent), Zone 2 (56 percent), and Zone 3 (59 percent), whereas in Zone 4, agriculture employs only a little under 40 percent of individuals. The industrial sector employs the smallest share across agro-ecological zones, especially in Burkina Faso. Variation across zones in the industrial sector’s share of employment is smallest in Côte d’Ivoire, suggesting a better geographic spread of industry across the country. In Burkina Faso, there are visible—though still only small—shares of industrial sector employment in Zones 2 and 4, with a share under 10 percent. Within Togo, Zone 1 has the smallest share employed in the industrial sector (8 percent), while Zone 4 has the largest share (47 percent). Within Benin, the pattern is similar, with Zones 44 1 and 2 having the smallest share (under 10 percent), and Zone 4 having the largest share (23 percent). In Côte d’Ivoire, variation across sectors in the share of the population employed in the industrial sector is smaller, while the share of industry is equally substantial in Zones 2 and 4, both being close to a 20 percent share of employment in industry. Within zones, those employed in agriculture are evenly spread. In locations where there are agglomerations, those people are more likely to be employed in industry and services. Agricultural Productivity While the agriculture sector faces various challenges and goals, enhancing agricultural productivity is paramount for both poverty reduction and economic growth and thus central to agriculture policy objectives. Growth in the agriculture sector is at least three times more effective in reducing poverty than growth originating in the rest of the economy (Ligon and Sadoulet, 2007). Particularly in Sub-Saharan Africa, leveraging agriculture-led growth requires a productivity revolution for smallholder farmers (World Bank, 2008). Globally, a 1 percent increase in yields is associated with a 0.91 percent decrease in the percentage of the poor population (Irz et al., 2001). However, it is worth noting that relative to other regions, both yields and poverty have changed least in Sub-Saharan Africa (De Janvry and Sadoulet, 2010; World Bank, 2008). For each of the countries in this sub-region, cereal yields were lower than the world average, even though Côte d’Ivoire clearly outperformed the other three countries. In 2014, the world average for cereal yields was 3,886 kg/hectare, while it was only 1,460 kg/hectare in Benin, 1,226 kg/hectare in Burkina Faso, 1,146 kg/hectare in Togo, and 3,254 kg/hectare in Côte d’Ivoire. From 2010 to 2014, Côte d’Ivoire saw a high increase in cereal yields (93 percent), whereas Togo had a minimal increase (8 percent). For the same period, cereal yields for Benin and Burkina Faso increased by 32 and 43 percent, respectively (World Bank, 2017). In this section, the major crops of each country are first summarized, comparing crops grown by the poor versus the non-poor. Then, in an effort to understand variation in productivity within each country, yields for a representative food crop (maize) and cash crops are mapped out across agro-ecological zones. 45 Box 4.1: Summary of the agriculture sector based on various world bank documents (Project Information Document [PID], Project Appraisal Document [PAD], and Systematic Country Diagnostic [SCD]) This inset provides brief background on the agriculture sector in each country to demonstrate that agricultural productivity, among other aspects, is central to the agenda. In Côte d’Ivoire, despite higher productivity relative to the other three countries, the lack of agricultural development and diversification and of structural transformation into agro-business are policy issues to which the country’s overall poor performance can be attributed. Inclusive growth can be achieved at least partially by developing the agriculture sector through enhanced productivity and developing agro-business and non-agro- business sectors. Despite its importance to the economy, the agriculture sector has had only a modest impact on income growth and poverty reduction in rural areas as the sector is characterized by low and unstable value addition. For example, while Côte d’Ivoire is the world’s largest exporter of cacao, this crop has seen declining productivity, and cacao farmers are falling into poverty. Other key cash crops include rubber, palm oil, cotton, and cashew nuts (Côte d’Ivoire SCD, 2015; Agriculture Sector Support Project PAD, 2013). In Burkina Faso, the lack of opportunities for rural populations is due to limited productivity gains in the agriculture sector, weak diversification, and slow emigration toward cities as a result of weak job prospects. A salient feature of Burkina Faso agriculture is that the increase in cereal production is entirely explained by the extension of harvested areas, while yields declined by about 3 percent. The agriculture sector consists mostly of subsistence farming, apart from cotton production, which accounts for roughly a third of exports. Burkina Faso also has a comparative advantage in mangoes, sesame, and shea nut, but is currently exporting at low levels in the value chain for such products (Burkina Faso SCD, 2017; Third Phase Community Based Rural Development, PID, 2012; Additional Financing for Agricultural Diversification and Market Development PID, 2014). In Togo, one of the pathways toward inclusive growth is the transformation of agriculture toward more productive, higher-value, and sustainable smallholder and commercial production. Despite a comparative advantage in agriculture, Togo has been unable to raise productivity consistently in the sector, including through diversification into higher value-added products. Agriculture is mainly of the subsistence type, and has seen a deterioration in the performance of the main export crops: cotton, coffee, and cacao. Togo has begun to address some of the constraints that affect the sector by recently adopting a plan aiming to reform the fertilizer subsidy program (Togo SCD, 2016). In Benin, the government strategy for the agriculture sector has been to improve productivity and strengthen diversification. Several challenges plague the sector. One of these is the reliance on a limited number of agricultural products, with most public resources concentrated on a single crop, cotton, which accounts for 25 –40 percent of GDP. Another challenge is the low level of both productivity and production mainly due to a lack of access to resources, improved technologies, and sources of finance. Notably, agricultural production systems rely on extensification and family labor, with limited use of improved inputs (Agricultural Productivity and Diversification Project PID, 2010; Nutrition Sensitive Agriculture & Capacity Building for Small Farmers PID, 2016). The above background highlights the policy attention placed on low agricultural productivity, with issues such as diversification and land extensification also deserving mention. Major Crops As a glimpse into the agriculture sector for each country, we look at which crops are grown by individuals engaged in agriculture. We use this as an indicator of which crops are important in the sense that more individuals engaged in agriculture rely on such crops. Thus, the focus is on crops grown by most people as opposed to crops that employ the most wage workers, take up more land, show higher output, or have the most value. The choice to focus on number of people growing a crop is largely due to data constraints. Nevertheless, this choice of focus also provides an opportunity to understand the agriculture sector from the smallholder angle as opposed to an understanding driven by large farming activity. We observe that both across and within countries, there is a great deal of variation in the types of crops grown. 46 Figure 4.2: Crops grown by poor and non-poor Côte d’Ivoire Burkina Faso Togo 47 Benin Source: Authors' calculations using Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. In Côte d’Ivoire, the crops produced by most people are a combination of food crops and cash crops. Half of the population grows cacao, followed by yam (29 percent), maize (21 percent), and cashew (18 percent). While cacao and cashew are largely sold in the market, only 20 percent of those that plant yam and 35 percent of those that plant maize sell some of those crops. In Burkina Faso, Togo, and Benin, the crops produced by most people are food crops. However, there seems to be more market exchange of crops in Togo than in Burkina Faso. In Burkina Faso, the crops planted by most are sorghum (77 percent), mil (55 percent), and maize (54 percent). Cash crops, such as cotton and sesame, are grown by only 15 percent and 29 percent, respectively, of those working in agriculture. In Togo, the most commonly produced food crops are maize (93 percent), cowpeas (57 percent), and sorghum (41 percent). Interestingly, there seems to be more market exchange of food crops in Togo than in Burkina Faso. For example, of those who plant maize in Burkina Faso, 20 percent sell some of their crop, whereas in Togo, 45 percent do so. Of those that plant sorghum in Burkina Faso, only 10 percent sell some of their sorghum crop, whereas in Togo, 57 percent do so. Much like Burkina Faso and Togo, Benin has most of its agricultural population producing some cereals (93 percent). With respect to cash crops, about 26 percent of those engaged in agriculture in Benin produce palm oil, and 23 percent produce cotton. 48 Overall, the crops produced by the poor are very similar to those produced by the non-poor. Similar proportions of the poor and non-poor grow key cash crops. However, there are distinctions in the food crops grown by the poor and the non-poor. In Côte d’Ivoire, compared to the non-poor, slightly more of the poor produce maize and fluvial rice and slightly fewer produce cacao. In Burkina Faso, slightly more of the poor produce sorghum, mil, cowpeas, and groundnuts. In Togo, there are more pronounced differences between crops grown by the poor and the non-poor, with many more of the poor producing crops such as cowpeas, yam, sorghum, gombo, soya, and rice. Cotton production spans the sub-region, particularly in a belt covering the southern parts of Burkina Faso as well as the northern parts of Côte d’Ivoire, Togo, and Benin. Interestingly, this cotton belt overlaps with areas of higher agglomeration that are not close to the capital city of each country. In Benin, cotton is more spread out throughout the country in terms of proportion of individuals planting it, being highest in the north and decreasing going south. Map 4.2: Cash crops across agro-ecological zones Source: Authors' calculations using Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. In agro-ecological zones where cotton is less prevalent, other cash crops come into play. This is true in all countries except in Burkina Faso, where Zone 4 has the highest proportion of both cotton and sesame production, while in Togo, we see soybean, in Benin palm oil, and in Côte d’Ivoire cacao. In Togo, while Zone 1 dominates in cotton prevalence, Zones 2, 3, and 4 dominate in soybean production, all of which have over 35 percent of those engaged in agriculture growing the crop. In Benin, there is an interesting and clear pattern whereby cotton production decreases going from North to South (Zone 1 to Zone 4) while palm oil production increases. In Zone 4, over 50 percent of those in agriculture produce palm oil, while only 3 percent grow cotton. In Côte d’Ivoire, cotton is grown only in Zone 2, while cacao production is most prevalent in Zones 3 and 4, with both zones having 70 percent of those working in agriculture growing it. 49 In choosing what other cash crop to observe for each country, we considered three factors. First, we considered the proportion of people growing the crop to ensure that we had a reasonable sample size that would allow us to observe differences across zones. Second, we considered the proportion of people growing the crop who sold some of that crop in order to verify whether the crop is in fact a marketable crop. Finally, we considered which crops are known to be cash crops in that country. Côte d’Ivoire shows uniquely diversified production of cash crops across zones: cashew in Zones 1 and 2, cotton in Zone 2, and cacao and some coffee in Zones 3 and 4. About 45 percent of those engaged in agriculture in Zone 1 and 42 percent in Zone 2 grow cashew, while fewer than 5 percent in Zones 3 and 4 grow the crop. Close to 50 percent in Zone 2 grow cotton, while in all other zones, cotton is non-existent. While 70 percent of those in Zones 3 and 4 grow cacao, only 14 percent in Zone 1 and 7 percent in Zone 2 grow the crop. Coffee is grown mostly in Zones 3 (20 percent) and 4 (13 percent) and by far fewer individuals in Zones 1 (5 percent) and 2 (2 percent). Map 4.3: Cash crops across Côte d’Ivoire’s agro-ecological zones Source: Authors' calculations using Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. The type of food crops grown also varies across agro-ecological zones, with maize production noted across all countries but to varying degrees. In Côte d’Ivoire, the top crops in Zones 1 and 2 are yam and maize, respectively. Maize is less prevalent in Zones 3 and 4, where only 11 and 12 percent, respectively, produce the crop. In Burkina Faso, maize is the most commonly produced crop in Zones 3 and 4 and is produced by many in other zones. Fully 20 percent of individuals in Zone 1 and 36 percent in Zone 2 produce maize. Sorghum is produced by many farmers across all four zones of Burkina Faso and is one of the top two crops in all zones, while mil is particularly important in Zones 1 and 2. In Togo, maize is produced by almost all of those involved in agriculture in all zones, while sorghum production is more prevalent in Zones 1 and 2 than in Zones 3 and 4 (see Appendix D). Food Crop Yields To explore productivity in food crop production, we look at yields, or the ratio of output (measured in kilograms) to land (measured in hectares). We focus on yields from one food crop, maize, because this food 50 crop is fairly prevalent across all four countries and is found in all zones in each country. This allows for a degree of comparability between yields across zones as well as across countries. However, we caution that maize is less prevalent in Côte d’Ivoire, where it only largely appears in Zone 3. Data constraints restrict us from being able to aggregate different cereal grains, for example, to obtain a measure of cereal yields. Map 4.4: Maize yields across agro-ecological zones Source: Authors' calculations using Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. In Côte d’Ivoire, maize yields in its lowest yield zone (1,603 kg/ha in Zone 1) are higher than in highest yield zones in any of the other three countries: 1,234 kg/ha in Zone 4 in Burkina Faso, 1,477 kg/ha in Zone 2 in Togo, and 1,582 kg/ha in Zone 2 in Benin. This highlights the fact that productivity in the agriculture sector in Côte d’Ivoire is very different from that in the other three countries. Across agro-ecological zones within Côte d’Ivoire, maize yields are much higher in Zones 3 and 4, the same zones where there is prevalent cacao production. Maize yields in Zone 4 are over five times higher than maize yields in Zones 1 and 2, even though maize production is most prevalent in Zone 2. Turning to yam yields in Côte d’Ivoire, we note that this crop is not widely produced in some zones. However, the yield for yam does not follow the same pattern, being highest for Zones 2 and 4 and lowest for Zones 1 and 3. While over time (2008–2013), cotton, cashew, and rice yields have been increasing, there has been a decrease in yields for important food crops such as roots and tubers. Moving forward, there needs to be an emphasis on improving food crop yields as these are more important for the northern areas of the country, where poverty rates are higher. Beyond this difference in food and cash crop yields, there are also large gaps in productivity for a given crop across households, suggesting scope for productivity gains (Christiaensen and Lawin, 2017). 51 In Burkina Faso, maize yields are highest in Zone 4, where conditions are generally favorable. Zone 4 thus sees relatively vibrant agriculture, with strong cotton production and relatively high maize yields. Despite the fact that conditions are also favorable in Zone 3, maize yields are about half in Zone 3 (683 kg/ha) of those in Zone 4, suggesting large variation in maize yields in Burkina Faso. Note that yields for sorghum do not follow the same pattern, being highest in Zone 2 and mostly similar across all the other zones. In Togo, Zone 3 has the lowest maize yields (817 kg/ha), while Zones 2 and 4 have the highest. Note that the disparity in maize yields between low-yield Zone 3 and high-yield Zone 4 is also quite large. In Benin, differences in productivity across zones are relatively low compared to the other countries, even though this could be at least partially a reflection of the fact that the Benin data on yields do not come from household surveys but rather from administrative data at the commune level from the Ministry of Agriculture. Thus, there is less comparability between Benin and the three other countries in the sub-region. Maize yields are lowest in Zone 4 (1,074 kg/ha), where agro-ecological conditions should be favorable but also where there is no cotton production. Maize yields are highest in Zone 2 (1,582 kg/ha), where there is cotton production. In Togo, Zone 2 is similar to Benin in having the highest maize yields in the country. However, unlike Benin, Togo has higher maize yields in Zone 4 than in Zone 3. One emerging pattern is that food crop yields are highest in those zones where key cash crops are prevalent. Cash crop production can contribute to food crop productivity through various pathways. First, cash crop income can help households overcome credit constraints on purchasing fertilizer and other inputs. Second, participation in a resource-providing scheme provides access to inputs via the marketing firm, and these inputs can in part be used for food crops. Third, cash crop income can allow households to make investments in tractors or draught animals, thereby increasing food crop productivity. Fourth, technical training provided by the marketing firm can increase food crop productivity (Schneider and Gugerty, 2011; Strasberg et al., 1999; and Minten, Randrianarison, and Swinnen, 2009). In Mali, for example, farmers take the credit and fertilizer extended to them by the cotton company and use it on their food crops (Thériault et al., 2015). The spillover effects of cash crop production or commercialization of agriculture on food crop productivity has been found in other settings, such as for cotton in Mozambique and coffee and sugarcane in Kenya (Govereh, Jayne, and Nyoro, 1999), cotton in Zimbabwe (Govereh and Jayne, 2003), and coffee in Kenya (Strasberg et al., 1999). Positive spillovers have also been found at both the household and the regional level in Zimbabwe (Govereh and Jayne, 2003). However, at least in Kenya, the effect of cash crop production on food crop yields varies by type of cash crop and geographical region, suggesting that policies on agriculture commercialization should carefully consider the context. Cash Crop Yields We find little variation in cash crop yields across zones in Côte d’Ivoire and Burkina Faso. However, there is considerable variation in cotton yields in Benin, with higher yields in Zone 1. We look at crop yields for the most prevalent cash crop in each country in order to compare productivity in cash crop production across agro-ecological zones within countries. We focus on cacao yields in Côte d’Ivoire and on cotton yields in Benin and Burkina Faso. However, cotton is not grown in Zone 1 in Burkina Faso and is therefore eliminated from this comparison. Due to the small sample size and low prevalence of cash crop production (soya or cotton), Togo is also eliminated from this analysis. It will be difficult to make any meaningful conjectures about differences in cash crop production across agro-ecological zones in Togo. 52 Map 4.5: Cash crop yields across agro-ecological zones Source: Authors' calculations using Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. In Côte d’Ivoire, there are no large differences across agro-ecological zones as regards cash crop productivity where cacao is prevalent. Zone 2 has lower yields of cacao (166 kg/ha), and this is also the zone where cacao is less prevalent. In Zone 2, where cotton is prevalent, average cotton yields (956 kg/ha) are similar to those in Burkina Faso and Benin. In Burkina Faso, there are also no differences in cotton yields across agro-ecological zones despite the fact that cotton is most prevalent in Zone 4. In Benin, there is greater variation in cotton yields across zones, with Zone 1 having the highest yields (940 kg/ha) and Zone 4 having half the cotton yields of Zone 1. Of the top two cotton-producing zones in Benin, Zone 1 has higher yields than Zone 2. Zone 1 in Benin has similar cotton yields to Burkina Faso’s (close to 1,000 kg/ha), while Zone 2 in Benin has lower crop yields. Assets, Inputs, and Output Markets Enhancing agricultural productivity is central to growth and poverty reduction. While agro-ecological conditions matter for productivity, other factors, such as market access, are also crucial. Various other factors must be in place to build productivity and ensure that it translates into development and poverty reduction (World Bank, 2008). One set of key ingredients involves increasing assets such as land, water, and human capital. The lack of assets is greatest in Sub-Saharan Africa, where farm size in densely populated areas is falling and investment in irrigation is negligible. Land markets can raise productivity, help households diversify income, and facilitate exit from agriculture, even if insecure property rights and poor contract enforcement restrict land markets. Although access to water and irrigation is a major determinant of productivity and stability of yields, in Sub-Saharan Africa, only 4 percent of the area under production is irrigated (World Bank, 2008). 53 Another set of key ingredients relates to improving access to both input and output markets. Market failures in Sub-Saharan Africa continue to be pervasive in Sub-Saharan Africa because of high transaction costs, risks, and limited economies of scale. Low fertilizer use remains one of the major constraints to agricultural productivity in Sub-Saharan Africa. However, improved productivity produces meager benefits if smallholders cannot sell their produce. Improving access to food staples markets and allowing smallholders to participate in the production of traditional exports can promote faster growth and benefit the poor (World Bank, 2008). In this section, we map access to assets, inputs, and output markets across agro-ecological zones in the sub- region. Specifically, we take stock of the use of fertilizer and phytosanitary products, land use, the quality of tenure security, and sales of agricultural produce. Identifying which areas are weak in terms of access to assets, inputs, and output markets is more useful from a policy perspective than learning that various agro- ecological zones have lower productivity simply because they are naturally disadvantaged. Moreover, to cite the case of Côte d’Ivoire, land tenure security, access to irrigation, and use of fertilizer and pesticides increase the likelihood of commercializing agriculture produce (Christiaensen and Lawin, 2017). The variation in yields we observe across agro-ecological zones lends itself to the idea that some areas have lower yields because they have fundamentally poorer growing conditions and are subject to poorer soil quality, vegetation, and rainfall. However, this is not necessarily the only reason for differences in yields. In fact, we find that some poorly endowed zones have reasonable maize yields, such as Zone 1 in Burkina Faso, and that some better endowed zones have poorer maize yields, such as Zone 4 in Benin. Prior work also suggests that cross-country differences in agricultural yields are more likely explained by economic decisions than by differences in agro-ecological endowments (Adampoulos and Restuccia, 2017). Instead, these economic decisions are influenced by institutions, constraints, and policies. Inputs to Production We now look at three variables regarding inputs to production across zones: spending on fertilizer, use of pesticide, and use of irrigation. Due to data constraints, the use of pesticide is not recorded in Benin, and the use of irrigation is not recorded in Burkina Faso and Togo. Map 4.6: Use of inputs and farm land across agro-ecological zones (a) Fertilizer spending (b) Pesticide use 54 (c) Irrigation use (d) Farm land size Source: Authors' calculations using Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. In Côte d’Ivoire, the use of pesticide and average spending on fertilizer both follow the same pattern across agro-ecological zones, where it is highest in the cotton producing Zone 2. Input use is substantially larger in Zone 2, where cotton is produced, than in Zones 3 and 4, where cacao is widely produced, and it is lowest in Zone 1. For example, the average amount spent on fertilizer is about US$620 (at 2011 PPP) in Zones 2 and 4, with the next highest spending on fertilizer reaching only US$100 on average. This difference in the use of inputs could also explain the observation that the highest yam yields are found in Zone 2 and the lowest yam yields in Zone 1. However, the vast difference in fertilizer spending is more likely due to cotton production in Zone 2. In Burkina Faso, there are equally pronounced differences across agro-ecological zones in the use of pesticide and average spending on fertilizer, where it is highest in the cotton producing Zone 4 and lowest or even negligible in Zone 1. For example, in Zone 4, average spending on fertilizer is about US$580 (at 2011 PPP), while in Zone 3, which also has some cotton production, spending on fertilizer is only US$130. There are also large differences in pesticide use across agro-ecological zones. These observed differences in input use could explain the high maize yields in Zone 4. However, it does not necessarily explain why maize yields are lowest in Zone 3 because this zone has slightly more input use than Zones 1 and 2. Compared to Côte d’Ivoire and Burkina Faso, Togo has low pesticide use across all zones. However, similar to both Côte d’Ivoire and Burkina Faso, fertilizer spending in Togo is highest in cotton producing zones. Benin also has high fertilizer spending in Zones 1 and 2. In Togo, particularly in Zone 1, US$80 (at 2011 PPP) is spent on fertilizer on average, and US$50 in Zone 2. Fertilizer spending is lowest to negligible in Zone 4. Thus, in Togo, the pattern whereby maize yields are highest in Zones 2 and 4 does not match the pattern for fertilizer spending. In Benin, the highest spending on fertilizer is in Zones 1 (US$110) and 2 (US$140), where cotton is produced, and spending is only about half that in Zone 4 (US$55). Comparing Zones 3 and 4, fertilizer spending is higher in Zone 3 than in Zone 4, which could partially explain the low maize and cotton yields in Zone 4. Land Use Not only do the cotton producing zones have high input use, farm land areas are also larger on average in these zones. We looked at average farm size among those who engage in agriculture. In Côte d’Ivoire, not only is irrigation, fertilizer, and pesticide use highest in the cotton producing Zone 2, 55 average farm land size is also highest (8.4 hectares), or double that in the other zones. The same is true for Burkina Faso, where farm size is largest for cotton producing Zone 4 (6.3 hectares). Interestingly, farm size in Zone 3, which has some cotton production, is almost half that in Zone 4. In Benin and Togo, Zone 4, which has favorable agro-ecological conditions, has smaller farm sizes on average. In Benin, farm size is highest in Zones 1 and 2 (about 5 hectares), and in Togo, farm size is highest in Zones 1 and 3 (also about 5 hectares). Data on irrigation are available only for Côte d’Ivoire and Benin. Use of irrigation in Benin is quite low, about only one-fifth of such use in Côte d’Ivoire, where irrigation is lowest in Zone 1 and highest in Zone 2. In Benin, irrigation is highest in Zone 1. The general pattern emerges whereby cotton growing areas use more pesticide, spend much larger amounts on fertilizer, and have larger farms. We see this in Zone 2 in Côte d’Ivoire (which also uses more irrigation) and in Zone 4 in Burkina Faso. Zones with favorable agro-ecological conditions such as Zone 4 in Benin and Togo tend to use the least amounts of fertilizer. However, we should note that zones with the poorest agro-ecological conditions, such as Zone 1 in Burkina Faso, also spend the least on fertilizer. Land Tenure Security In rural West Africa, the allocation and enforcement of land rights mostly operates through a diverse and overlapping set of customary arrangements at the village or local level. Increasing pressure on natural resources and the absence of written documentation regarding land use have given rise to land conflicts over inheritance as well as disputes between villages, farmers, and pastoralists. This lack of formal land rights may lead to under-investment and sub-optimal yields. In theory, the codification of private property rights within an effective legal framework should increase agricultural investment and productivity. In this sense, land reform is a critical policy action in the sub-region, as it is elsewhere. Box 4.2: Background on land reform in Benin, Côte d’Ivoire, and Burkina Faso We first provide a brief background on land reform in some of the countries in the sub-region before mapping out land tenure security across agro-ecological zones. In Benin, a Rural Land Use Plan (Plan Foncier Rural, PFR) has been in operation since 1993 and has been recognized by the 2007 Rural Land Act. It is considered different from more standard land formalization programs in two respects. First, it recognizes that existing customary arrangements provide legitimate claims to property that can be formalized. Second, it sets up a decentralized procedure for the establishment of formal property rights. Recent work suggests that the demarcation process of PFRs lead to long-term investments in agriculture (Goldstein et al., 2015). In Côte d’Ivoire, insecurity of land tenure has been one of the root causes of conflict and constrains investment in agricultural development and agro-business. Almost all farmland is owned and transferred according to customary law. Because customary land tenure systems are not well defined or consistently applied, their use has led to conflict. The 1998 land reform legislation, which aimed at moving from customary land tenure to more modern arrangements, has been slow in its implementation (Côte d’Ivoire SCD, 2015; Agricultural Sector Support Project, PAD, 2013). In Burkina Faso, despite land reform carried out by the government through a consultative process over the last few years and the consequent adoption of various legislation, its enactment has been slow. Only a small fraction of all land has been registered formally. Traditional laws for land management and community-based ownership continue to hamper land transactions as their relationships with modern laws have not yet been fully clarified (Burkina Faso SCD, 2017). 56 Map 4.7: Land tenure security across agro-ecological zones Source: Authors' calculations using Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. We first look at reported ownership of land across agro-ecological zones. Across all countries, there is relatively high reporting of owning at least one plot, except in Benin. In Benin, while 76 percent of individuals own at least one plot in Zone 4, fewer than 60 percent own at least one plot in all other zones. In Burkina Faso and Côte d’Ivoire, compared to other agro-ecological zones within each country, there is only slightly less ownership of land in the cotton producing zones, particularly Zone 4 in Burkina Faso (70 percent) and Zone 2 in Côte d’Ivoire (74 percent). Among those who own land, we look at the securitization of land tenure across agro-ecological zones. There are only slight differences in the definition of securitization across countries. In Côte d’Ivoire and Togo, weak securitization is defined as having no form of documentation for at least one plot owned. In Burkina Faso and Benin, weak securitization is defined as having no form of documentation or verbal agreement for at least one plot owned. Thus, our definition of “securitization” is less stringent for Burkina Faso and Benin (which includes verbal agreements for tenure security) than for Côte d’Ivoire and Togo. Across countries, there seems to be weaker securitization of land in higher value agricultural areas, particularly in cotton producing zones. Among those who own some land, there is better securitization of ownership in Zone 4 in both Benin and Togo. In Côte d’Ivoire, compared to Zones 1 and 2, there are more individuals with better securitization of land in Zones 3 and 4. In Burkina Faso, where ownership is low, securitization of land is also lower in Zone 4. Access to Output Markets Achieving reductions in poverty via enhanced productivity and improved input and land markets is unlikely to be achieved if households are unable to access markets to sell their produce. The lack of access to markets inhibits the poor from participating in the benefits of improved agricultural productivity (Schneider and Gugerty, 2011). In this section, we map out the sales of crops across agro-ecological zones. Due to data constraints, data on the sale of crops is not available in Benin. 57 Map 4.8: Sale of agricultural produce (a) Sell some of their crop (b) Revenue from crop sales Source: Authors' calculations using Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. In Côte d’Ivoire, there is not much variation across zones in the proportion of individuals who sell some of their crop, even though the value of sales is substantially higher in Zone 2. In Zones 1 and 2, about 73 percent of individuals sell some of their crop compared to about 82 percent in Zone 4. However, among those who sell some of their crop, the value of sales varies widely across zones. Zone 2 has the highest average value of crop sales (about US$3,800 at 2011 PPP), followed by Zone 4 (US$2,700). The average value of crop sales in Zone 1 is less than one-third of the average value of crop sales in Zone 2, and that in Zone 3 is only about half of the value of sales in Zone 2. In Burkina Faso, very few sell their crop in Zone 1, while many more do so in Zone 4, where the value of sales is also substantially higher. There is a great deal of variation in access to crop markets in Burkina Faso, where only 30 percent of individuals in Zone 1 sell some of their crop, whereas 73 percent do so in Zone 4. Among those that sell some of their crop, the average value of crop sales is about four times higher in Zone 4 (US$2,200 at 2011 PPP) than in Zones 1 and 2. Moreover, the average value of crop sales is also over three times higher in Zone 4 than in Zone 3. In Togo, both the proportion of individuals who sell some of their crop and the value of crop sales among those who do so is highest in Zones 1 and 3. In Zone 1, 86 percent sell some of their crop, while in Zone 3, 79 percent do so. Compared to Burkina Faso, there is greater market exchange of crops in Togo. In Togo, the zones with the highest proportion of individuals who sell some of their crop are those zones with lower maize yields. This is unsurprising as maize is primarily a food crop. Thus, these zones are likely to be selling crops other than maize. Among those who sell some of their crop, the average value of crop sales is much higher in Zone 1 (US$1,500 at 2001 PPP) and 3 (US$1,100) than in Zones 2 (US$600) and 4 (US$700). 58 Chapter 5: Putting It All Together Having graphically established a two-way relationship between welfare and each of the three building blocks of the economic geography literature discussed in Chapter 3, and carried out a separate process for agricultural productivity in Chapter 4, we now put all the factors together. Specifically, in this chapter, we aim to quantify the relationship between welfare and agricultural productivity. Moreover, we seek to analyze correlates of variation in both welfare and agricultural productivity across space. To do so, we define three sets of explanatory variables corresponding to the three building blocks of the economic geography: natural endowment, agglomeration, and market access. Our dependent variables are welfare and agricultural productivity. Welfare levels are measured by average household consumption in each administrative unit.18 Hereafter, we use consumption and income interchangeably. While different cash crops and food crops are harvested across space (see Figure C. in Appendix C for a complete list of crops), data on maize yields (in kg/ha) are available across all four countries and in most administrative units and are therefore used here as a proxy for agricultural productivity. In this paper, we provide only a static picture of the relationship between welfare, agricultural productivity, and the three sets of explanatory variables. The results should therefore be interpreted as correlates only. There has been an ongoing debate in the economic geography literature about whether a location’s levels of per capita income and other economic dimensions are determined by geographical and ecological variables. Many researchers have provided evidence supporting the view that such links are strong (see, for example, Gallup et al., 1999 Sachs, 2000; Gallup and Sachs, 2001; Sachs and Malaney, 2002), while others have argued that the role of geography in explaining spatial patterns of per capita income operates through various direct channels (e.g., productivity and trade) or indirect channels (e.g., choice of political and economic institutions) with little direct effect of geography on income (see, for example, Acemoglu, Johnson, and Robinson, 2001; Easterly and Levine, 2002; Rodrik, Subramanian, and Trebbi, 2004). How does this play out in the case of Benin, Burkina Faso, Côte d’Ivoire, and Togo? The answers are: (a) A region’s income levels and its agricultural productivity are closely related in the four countries studied; (b) Except for being coastal or landlocked, geographical covariates do not appear to be directly associated with a region’s per capita income. However, the relationship between geography and welfare is mediated by agglomeration economies and market access; (c) As expected, there is a strong link between agro-ecological characteristics and yields per hectare; (d) What is notable is the persistence of the correlation between geography and crop yields regardless of whether population density, market access, or farm inputs are controlled for. This pattern should be taken into consideration when planning development strategies in the agriculture sector. Relationship between Welfare and Agricultural Productivity We first show that agricultural productivity is positively correlated with per capita expenditure. This rationalizes the geography of agricultural productivity as a means to better understand the geography of poverty. For Benin, Burkina Faso, Cote d’Ivoire, and Togo, we find that on average, a 10 percent increase in mean maize yields is associated with a 1.7 percent increase in mean per capita expenditure (Figure 5.1). 18 Administrative units are communes in Benin, provinces in Burkina Faso, departments in Côte d’Ivoire, and prefectures in Togo. 59 Figure 5.1: Correlation between poverty and agricultural productivity β= 0.17*** (including country fixed effects) Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. The relationship between higher productivity and lower poverty in the four countries is consistent with what has been observed in other countries. Globally, there is an almost one-to-one relationship between crop yields and poverty alleviation such that a 1 percent increase in agricultural productivity is correlated with a 0.9 percent reduction in poverty (Irz et al., 2001). In India, the poor have gained in both absolute and relative terms from increased farm yields (Datt and Ravallion, 1998). In Madagascar, higher yields are linked to improvements in food security (Minten and Barrett, 2008). In Ethiopia, more complex measures of agricultural productivity also correspond to reductions in poverty (Abro, Alemu, and Hanjra, 2014). There are multiple pathways from farm productivity to poverty reduction. Beyond the potential direct effects on farmers, there are also indirect effects such as non-farm job creation as well as linkages to the rest of the economy. Improvements in productivity can lower food prices (which benefits poor net buyers of food) and raise the wages of unskilled workers (which benefits poor unskilled workers). However, food prices and wages may change slowly over time such that the longer-term effects of agricultural productivity on poverty reduction may outweigh the short-term effects (Schneider and Gugerty, 2011; De Janvry and Sadoulet, 2010; Minten and Barrett, 2008; Datt and Ravallion, 1998). Three Sets of Explanatory Variables for Three Building Blocks The next question is: Among the three building blocks of economic geography, which factors are significantly associated with a location’s welfare and agricultural productivity? We start with natural endowment covariates consisting of six continuous agro-ecological variables that arguably affect a region’s agricultural productivity and therefore income. These are temperature, precipitation, soil quality, elevation, latitude, and ruggedness.19 While high-elevation locations often have low market access due to poor transportation, once ruggedness and coastal locations are controlled for, it is best interpreted as an agro-ecological variable. We thus introduce a seventh variable, namely a coastal dummy, which takes a value of 1 if the administrative unit has a coast and zero otherwise. We then add an agglomeration economies variable measured as the log of the number of people per square kilometer. For 19 Two additional agro-ecological variables—length of growing period and slope—are strongly correlated with latitude and precipitation (> 0.9) and were therefore are excluded from our regressions. 60 our regressions on agricultural productivity, we also include the share of employed population in the agriculture sector. Finally, our covariate for market access is the log of market access index described in Chapter 2. Table 5.1 reports the statistics of the variables used. Table 5.1: Statistical summary Number of Standard Mean observations deviation Dependent variables Consumption per capita (2011 PPP) 265 893 355 Maize yields (kg/hectare) 237 1,918 2,823 Natural endowment covariates Temperature (celsius) 263 26.891 0.927 Precipitation (mm/year) 261 1,171 268 Soil 261 4.083 2.584 Latitude 261 8.628 2.314 Elevation (m) 261 229 119 Ruggedness 263 0.202 0.185 Coastal dummy 265 0.064 0.245 Agglomeration covariate Population density (people/km2) 265 244 893 Share of population in agriculture 264 0.605 0.272 Market access covariate Log (market access index) 265 9.825 2.036 Note: Observations are recorded at administrative unit levels: 77 communes in Benin, 45 provinces in Burkina Faso, 107 departments in Cote d'Ivoire, and 36 prefectures in Togo. Correlates of Welfare Table 5.2 presents the results of the multivariate regressions in which we relate an administrative unit’s per capita consumption level to the three sets of explanatory variables described above. To assess spatial differences within a country, we include country fixed effects in all of our regressions. Column 1 looks at the relationship between consumption and six agro-ecological characteristics. We are particularly interested in the relative effect of each characteristic on welfare. Column 2 adds the seventh variable of interest, a dummy variable that takes the value 1 if the administrative unit has a coast and zero otherwise. Column 3 introduces the agglomeration economies variable of population density. Finally, Column 4 adds the market access covariate. The first observation is that the closer a location to the equator, the higher its income , with the coefficient of the latitude variable being negative and significant (Column 1). More precisely, an increase of one degree of latitude is correlated with a 9.4 percent drop in per capita consumption. This confirms our story described in Chapter 3, namely that the North is generally poorer than the South. Keeping other factors constant, cooler temperatures and less-rugged terrain are also associated with higher income. However, rainfall levels, soil quality, and elevation do not seem to be directly linked to income. The second observation is that being on the coast signals wealth (Column 2). On average, per capita consumption in a coastal area is 55 percent higher than in a landlocked location. After controlling for whether or not a location is on the coast, the income pattern from North to South still holds. A one-degree increase in latitude is correlated with a 10.2 percent decrease in per capita consumption. Ruggedness 61 remains a key variable with a significant link to a location’s income, while the remaining f our natural endowment covariates (temperature, rainfall levels, soil quality, and elevation) do not. Interestingly, once population density is taken into account, none of the agro-ecological characteristics, except the coastal dummy, is significant (Column 3). In other words, a location’s wealth is solely related to its level of agglomeration and its situation near the sea. Between two locations with the exact same population density, the one on the coast is 17.5 percent richer than one located inland. Similarly, holding other variables constant, a 10 percent increase in population density is associated with a 1.6 percent increase in income. Table 5.2: Factors associated with spatial differences in poverty: coastal location, population density, and market access Regression results (1) (2) (3) (4) Dependent variable Log (consumption per capita) Natural endowment covariates Temperature (Celsius) −0.129* −0.079 −0.032 −0.028 (0.075) (0.072) (0.066) (0.065) Precipitation (mm/month) 0.000 −0.000 0.000 −0.000 (0.000) (0.000) (0.000) (0.000) Soil quality (per mille) −0.002 −0.013 −0.010 −0.007 (0.020) (0.018) (0.012) (0.012) Latitude (degrees) −0.094* −0.102** −0.039 −0.047 (0.051) (0.043) (0.035) (0.034) Elevation (m) −0.001 0.001 0.000 0.000 (0.001) (0.001) (0.000) (0.000) Ruggedness (in 100 m) −0.451*** −0.301* −0.149 −0.136 (0.169) (0.157) (0.134) (0.128) Coastal dummy 0.545*** 0.175** 0.210*** (0.088) (0.071) (0.068) Agglomeration covariate Log (population density) 0.157*** 0.092*** (0.024) (0.032) Market access covariate Log (market access index) 0.046*** (0.017) Country fixed effects Yes Yes Yes Yes Number of observations 261 261 261 261 Adjusted R squared 0.525 0.626 0.711 0.724 Note: Standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. A key message from Column 4, which includes all three sets of explanatory variables (natural endowment, agglomeration, and market access) is that all three sets matter for a region’s income in Benin, Burkina Faso, Côte d’Ivoire, and Togo. However, a region’s position in relation to the sea is the only geographic characteristic associated with that region’s wealth. With the same levels of agglomeration and market access, a coastal location’s income is 21 percent higher than that of a landlocked one. What these results suggest is that with the exception of being landlocked, the relationship between geography and welfare is mediated by population density and market access. In other words, areas 62 good for growth will either attract people or experience stronger population growth and at the same time receive investments in infrastructure. Thus, when controlling for population density and market access, we no longer detect any relationship between welfare and temperature, latitude, or ruggedness. As shown in Chapter 3, agglomeration and market access often go hand in hand (i.e. densely populated areas enjoy better market access and vice versa). Therefore, after controlling for market access levels, the relationship between agglomeration and income declines in magnitude, albeit remaining significant (Columns 3 and 4). What Column 4 implies is that keeping all variables constant, including market access, a 10 percent increase in population density correlates with a 0.9 percent increase in wealth compared to 1.6 percent in Column 3, where market access is not accounted for. In addition, a 10 percent improvement in market access is associated with a 0.5 percent increase in income. While the results suggest that human activity does respond to geography in the sense that people move or build roads where the environment is more favorable, we find that the benefits of being along the coast are not arbitraged away by migration or increased market access. This is an interesting finding in that it reveals the untapped potential for economic development provided by access to international trade for the three coastal countries (Benin, Côte d’Ivoire, and Togo). Correlates of Agricultural Productivity In addition to the three sets of covariates of natural endowment, agglomeration, and market access, this section explores the agriculture module found in the household consumption surveys in order to select farm inputs that matter for agricultural productivity. The challenge is to identify a set of farm input variables that are not only common across surveys in Benin, Burkina Faso, Côte d’Ivoire, and Togo but also available in most administrative units. The three variables that meet such requirements are: spending on fertilizer, land size, and share of farmers having weak land tenure security. Column 1 evaluates the correlation between agricultural productivity and agro-ecological characteristics, including the coastal dummy. Column 2 adds agglomeration variables consisting of population density and share of the employed population working in the agriculture sector. Column 3 introduces market access, and Column 4 adds three farm input variables. Our results in Column 1 show that 4 out of the 7 agro-ecological characteristics under study are important to agricultural productivity. These are temperature, latitude, elevation, and location in coastal areas. Similar to the relationship with income, temperature is also negatively correlated with agricultural productivity: on average, one additional degree Celsius is associated with an 87 percent decrease in agricultural productivity. Similarly, land located at high elevations has lower crop yields, with a 100-meter higher location correlating with a 70 percent reduction in production. Notably, being on the coast signals not only wealth but also high agricultural production. We find that a coastal region is associated with a more than 100 percent increase in yields. However, controlling for a region’s location (whether landlocked or coastal), the further a region from the equator, the higher its agricultural yield, while one degree of latitude toward the North correlates with a 44.6 percent increase in agricultural productivity. This relationship is opposite to that between latitude and income presented in the previous section (Table 5.2, Column 2), suggesting that economic activity near the equator (i.e. the South) must expand beyond the agriculture sector to make up for the decline in crop yields. This pattern is consistent with the distribution of sectoral employment analyzed in Chapter 4, which showed that the concentration of the employed population in the industry and service sectors increases toward the South (Table 5.3). 63 Table 5.3: Role of geographical differences in agricultural productivity, natural endowments (temperature, latitude, elevation, coastal location), and spending on fertilizer Regression results (1) (2) (3) (4) Dependent variable Log (maize yield) Natural endowment covariates Temperature (Celsius) −0.873*** −0.870*** −0.859*** −0.961*** (0.266) (0.269) (0.265) (0.317) Precipitation (mm/month) 0.001 0.001 0.001 −0.000 (0.001) (0.001) (0.001) (0.001) Soil quality (per mille) −0.048 −0.051 −0.049 −0.035 (0.044) (0.045) (0.045) (0.047) Latitude (degrees) 0.446*** 0.415*** 0.401*** 0.351** (0.120) (0.131) (0.132) (0.146) Elevation (m) −0.007*** −0.007*** −0.007*** −0.007*** (0.002) (0.002) (0.002) (0.002) Ruggedness (in 100 m) −0.576 −0.522 −0.505 −0.237 (0.487) (0.501) (0.496) (0.503) Coastal dummy 1.081*** 1.105*** 1.166*** 1.148*** (0.248) (0.306) (0.311) (0.301) Agglomeration covariate Log (population density) −0.085 −0.170 −0.097 (0.093) (0.111) (0.113) Share of population in agriculture −0.501 −0.358 −0.456 (0.373) (0.348) (0.339) Market access covariate Log (market access index) 0.074 0.040 (0.045) (0.045) Agricultural input covariates Log (fertilizer spending) 0.107*** (0.038) Log (land size) 0.115 (0.138) Land title dummy −0.188 (0.309) Country fixed effects Yes Yes Yes Yes Number of observations 233 232 232 218 Adjusted R squared 0.498 0.503 0.507 0.527 Note: Standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. In contrast to NEG literature suggesting that agglomeration economies and market access can help farmers take advantage of better prices and a wider selection of agricultural labor, farm inputs and technology, and better markets for harvested crops, we do not find such a link in the four countries 64 under study. The correlations between agglomeration covariates, market access index, and agricultural productivity are not significant (Columns 2 and 3). This finding suggests that we may in fact be dealing with two types of agriculture: a subsistence agriculture, whereby most crops are cultivated for home consumption and where investments are less sensitive to market access, and a commercial agriculture concentrated along the coastline and benefiting from higher investments in inputs (such as fertilizer). When we introduce farm input covariates into the model (Column 4), only the value of expenditure on fertilizer shows a strong correlation with crop yields. A 10 percent increase in fertilizer spending is associated with a 1.1 percent increase in agricultural productivity. However, land ownership regime and land size do not appear to play a significant role in yields above and beyond their potential influence through fertilizer use. An important pattern that emerges from the four regressions is the persistent link between natural endowment and crop yields. The sign and magnitude of the coefficients for temperature, latitude, elevation, and the coastal dummy barely change regardless of whether or not variables for agglomeration, market access, or farm inputs are taken into account. 65 Chapter 6: Policy Recommendations and Further Studies Over the past five years, GDP in Benin, Burkina Faso, Côte d’Ivoire, and Togo has grown at the impressive rate of approximately 5 percent annually. However, such economic growth has not yet translated into high levels of prosperity. Poverty rates remain relatively high, even by SSA standards. Moreover, spatial disparities in welfare and poverty are evident throughout all four countries, with a typical household in a leading area consuming as much as 7 times more than a similar household in a lagging area. Within countries, there is also strikingly wide variation in poverty rates across administrative units. As illustrated in the World Development Report – Reshaping Economic Geography (World Bank, 2009), within-country disparities can be a potential source of tensions between lagging and leading regions and may affect the country’s overall growth and political stability. How can our findings help policy makers reduce geographical differences in welfare while boosting growth? Based on our analysis, we propose four broad policy recommendations: (a) Urbanization: We find that many of the leading areas have not yet maximized the benefits of agglomeration economies, especially in Burkina Faso and Côte d’Ivoire. Based on NEG literature, there is scope for increasing the concentration of economic activities and labor in these areas to further take advantage of economies of scale and boost economic development. However, it is important to consider complementary policies to urbanization, including removing barriers to labor mobility so that people can migrate to leading areas where labor demand and productivity are higher, and investing in urban infrastructure and the provision of public services to accommodate a potentially larger number of users. (b) Increasing agricultural productivity: Not all rural families can move to urban locations. For those staying in the agricultural sector in rural areas, policy makers may consider improving their welfare by increasing agricultural productivity. Potential areas of improvement include land tenure, irrigation, use of farm inputs such as fertilizer, and research and development (R&D). Given that agro-ecological endowments seem to influence crop yields more than agglomeration and market access, there is a strong need for tailoring inputs and R&D to specific agro-ecological zones. (c) Fiscal transfers: There exist geographical pockets of poverty where the costs of reaching the poor are very high. These areas are characterized by a combination of high poverty rates and low poverty density. Another set of lagging areas with little prospect of growth consists of those with unfavorable agro-ecological characteristics and limited opportunities to diversify into non- agricultural sectors. Our findings imply that some lagging areas may not be able to improve their welfare after all. This may call for pro-poor fiscal transfers through a system of inter-regional transfers to ensure equity across the leading and lagging regions. Since there are limited job opportunities in low-density and lagging areas, it is important that local government use the fiscal transfers received to invest in portable assets (health and education) for its citizens so that they can join a healthy and educated workforce if they choose to migrate. (d) Safety net programs: Not all poor people, especially the vulnerable, can benefit from the policies proposed above. The need to maintain strong safety net programs targeting the poor and the vulnerable remains strong. New technologies such as e-vouchers and mobile transfers make it possible for such programs to reach targeted beneficiaries in low-density areas in a cost-effective way. Moreover, safety net programs should be part of an overarching poverty reduction strategy consisting of interacting with and working alongside urban policy, agricultural productivity boosting programs, and other policies aimed at eradicating poverty and reduce vulnerability. Urbanization We observe a strong link between agglomeration economies and income levels in the four countries. However, many of the leading regions still have low population density, i.e. fewer than 150 people per 66 square meter, especially in Burkina Faso and Côte d’Ivoire. Such low population density makes it difficult for these regions to reap the benefits of agglomeration productivity and thus further advance their economic development. Arguably, an urban agglomeration economy brings many economic benefits. The first and foremost advantage is a reduction in transportation costs for goods as producers are located near their customers. In the 1990s, New York and London were manufacturing powerhouses as factories were built in and around these cities for better access to customers and transportation services (Lall, Henderson, and Venables, 2017). Moreover, the advantage of agglomeration economies increases with scale. Rosenthal and Strange (2004) show that each doubling in city size increases productivity by 5 percent, and the elasticity of income with respect to city population is between 3 and 8 percent. Urbanization is strongly associated with productivity gains through their links to structural transformation and industrialization. As a country urbanizes, people move from rural to urban areas in search of better job opportunities in terms of higher pay or productivity. Thus, a complementary and necessary policy to favor urbanization is the removal of barriers to labor mobility so that people are able to not only physically migrate to urban cities but also to move to other economic sectors that offer better returns. A highly dense location (in population or firms) also brings down the costs of certain public investments such as infrastructure and basic public services. The average costs of such programs are lower when their users are many in number and densely grouped together. Input costs for firms located near each other also decline as firms share infrastructure and suppliers. In addition, thick labor markets allow firms to reduce search costs and to have access to a larger pool of potential workers. This reduction in mobility costs thus allows improved allocation efficiency. Close spatial proximity also promotes innovation and knowledge sharing among people and firms . International evidence illustrates how knowledge spillovers are a crucial element in improving the productivity of successful urban cities. East Asia’s success story (e.g., China, the Republic of Korea, Vietnam) shows a strong relationship between urbanization and economic development (Lall, Henderson, and Venables, 2017). To boost growth, policy makers may consider promoting urban planning in some leading regions . Policies that support urbanization may include formalizing land markets and making early and coordinated urban infrastructure and public service investments. A formal market for urban land offers buyers the legal protection of the government and generates the public good of accurate valuation. Not only is this a precondition for land consolidation, which converts low-density residential use into higher-density housing or clusters of new commercial structures, but it is also an incentive for farmers to invest in inputs such as fertilizer given that the risk of expropriation is now lower thanks to more secured property rights. In addition, the early installation of coordinated urban infrastructure helps shape urban structures and save costs. If postponed until after population settlement, such infrastructure is far more difficult and expensive to install. Another important policy for urbanization is the provision of public goods and services to ensure quality of life for an increasing urban population (Lall, Henderson, and Venables, 2017). Agricultural Productivity It is important to improve welfare in rural areas. Not all rural families can move to urban cities, for many reasons, including limited absorptive capacity of denser areas, government policies preventing slum proliferation, poorly defined land ownership, risk aversion, or poor information (Gollin, Kirchberger, and Lagakos, 2016; de Brauw, Mueller, and Lee, 2014). For those remaining in the agriculture sector in rural areas, policy makers may look to improve their productivity, which has been shown in our analysis to have a strong link with welfare. In this report, we illustrate how securing land tenure and irrigation are largely untapped in the sub- region. Furthermore, the use of inputs such as pesticide and fertilizer is low, which suggests the potential 67 for targeted farmer subsidies and the need for finance so as to enable access to credit for the purchase of inputs and machinery. Access to savings and insurance will also be crucial in allowing farmers to mitigate both idiosyncratic and covariate weather shocks. The World Bank’s World Development Report (WDR, 2009) recommends two broad sets of effective instruments for leveraging agriculture for development. The first is to increase access to assets such as land, water, education, and health. The second is to make smallholder farming more productive and sustainable. These instruments and goals are naturally intertwined as access to assets enhances productivity, and sustainable smallholder farming is a prerequisite to securing productivity in the longer term. More specifically, WDR 2009 emphasizes several possible instruments such as improving price incentives and increasing the quality and quantity of public investment, making output and input markets work more efficiently, improving access to financial services and reducing exposure to uninsured risks, enhancing the performance of producer organizations, and promoting innovation through science and technology. In Sub-Saharan Africa, market failures in input markets continue to be pervasive, and as a result, low fertilizer use is one of the major constraints on increasing agricultural productivity in the region. Although providing fertilizer subsidies is a clear strategy, the focus should shift toward more sustainable solutions such as targeted vouchers to enable farmers to purchase inputs and stimulate demand in private markets as well as providing matching grants to underwrite start-up costs of entry into markets for private distributors. Which instrument or combination of instruments will prove most effective in enhancing productivity will depend on the context, not only of each region or country but also within each region and for each value chain. Furthermore, innovations in information and communication technology (ICT) can be leveraged to make markets work better through interventions known collectively as “e-agriculture.” For example, both Benin and Burkina Faso have seen projects launched to collect and disseminate food market prices by short message service (SMS). In Côte d’Ivoire, a mobile phone program called CocoaLink connects farmers to agricultural experts who can address questions concerning fertilizer application or disease and pest control in real-time. Thus, e-agriculture can spread information on market prices and best farming practices. Fuglie and Rada (2012) compare a set of policies that may correlate with total factor productivity (TFP) growth across countries in Sub-Saharan Africa, namely investment in research (through national and international agricultural research centers such as Consultative Group for International Agricultural Research [CGIAR]), economic policies (commodity price interventions, trade tariffs, and input subsidies), human capital (education and health), infrastructure (extent of roads), and the prevention of armed conflicts. Although research may have long-lasting effects on productivity, that effect may be delayed. Fuglie and Rada find that for every US$1 spent on R&D, US$3–US$5 in benefits are generated. Higher returns to agricultural research in SSA could be seen through strengthening the CGIAR system, followed by strengthening national agricultural systems in larger countries. Beyond R&D investments that have direct effects on TFP growth, policies that strengthen the enabling environment are also crucial to raising agricultural productivity in the region. Economic policies that reduce net tax on the agriculture sector (i.e. subsidies) and increase the levels of labor force education are found to increase agricultural productivity. On the other hand, armed conflict and the spread of Human Immunodeficiency Virus/Acquired Immune Deficiency Syndrome (HIV/AIDS) are barriers to agricultural productivity. Given that agro-ecological endowments seem to influence agricultural yields more than agglomeration and market access, policies designed to improve agricultural productivity should be sensitive to these agro- ecological differences. R&D efforts and the available input mix (such as fertilizer blends and seed varieties) should be localized to agro-ecological zones. In Mali, for example, the same fertilizer mix that is suitable for cotton is inappropriately used country-wide despite the fact that such mix is less effective outside of cotton production. 68 Fiscal Transfers Our findings highlight the stylized fact that some lagging regions may be falling further behind as a result of the development process. Over time, economic geography will continue to favor economic concentration in leading areas and make it more difficult for poor areas to catch up. A set of locations that are particularly vulnerable to this development path are areas with not only high poverty rates but also low poverty densities. Such combination leads to extremely high cost of investments in infrastructure (e.g., market access, irrigation) and public services (e.g., electricity, water). Another possible set of vulnerable locations is characterized by unsuitable agro-ecological characteristics for agricultural production and a lack of job opportunities in non-agricultural sectors. Evidence from our analysis points to a persistent relationship between natural endowments and agricultural productivity regardless of the levels of agglomeration, market access, or farm inputs used. One option is to create opportunities for residents to move to higher-return areas should they choose to do so. Policies supporting this option are described in the section on urbanization above. However, not everyone wishes to move, for various reasons such as cultural barriers (e.g., ethnicity, language) or risk aversion. This calls for pro-poor fiscal transfers. Lagging areas have a low base of economic activity to tax and thus yield low revenues. This budget constraint prevents them from providing adequate social safety nets for poor residents, who represent a large share of the population. Moreover, it limits the ability to fund investments in human and physical capital and deliver public services, all of which are likely to have extremely high delivery costs per user. Achieving equity through fiscal transfers can therefore ensure a level playing field (World Bank, 2010). However, transferring financial resources to lagging regions alone may not be sufficient. This should be accompanied by improving capacity and accountability on the part of local government. Moreover, since there are limited job opportunities in low-density and lagging areas, it is important that local government use the fiscal transfers received to invest in portable assets (health and education) so that citizens can join a healthy and educated workforce if they choose to migrate (World Bank, 2010). Safety Net Programs There remains an urgent need to maintain strong safety net programs as the proposed policy above may not reach all those in need. Ultra-poor people may not have the financial means or adequate information to migrate to urban areas. Poor women may not have access to farming or other job opportunities. Poor children may not be able to attend school due to home–school distance or family financial constraints. Arguably, safety net programs can play crucial roles in development policy (Grosh et al., 2008). First, they aim to redistribute income to the poorest and the most vulnerable, resulting in an immediate impact on poverty and inequality. Second, safety nets can allow households to take up investments both in terms of human capital and financial investments with a view to securing their future. Third, such programs help the poor and vulnerable manage risks. Finally, they can free other sectors from the role of redistribution and concentrate on efficiency instead. However, it may be prohibitively expensive for these programs to reach the poor in some areas. As illustrated in our analysis, such areas are characterized by high poverty rates but also low poverty density. A general payment system would require distributing agencies such as NGOs, public agencies, banks, and retail stores to physically deliver cash, in-kind goods, or vouchers to beneficiaries in sparsely populated areas where transportation costs are high. Fortunately, innovative and affordable technology such as electronic and mobile vouchers allow such programs to cover low-density areas efficiently. For example, targeted beneficiaries can receive vouchers automatically on their phone. They can redeem these vouchers for cash or in-kind goods at any participating 69 retailers or agency at any time. This process not only saves delivery costs but also provides transparency and overcomes payment delays as well as complicated and inconvenient redemption systems (World Food Programme,, 2014). Limitations and Further Studies This report aims to provide policy makers with stylized facts about spatial differences in welfare, poverty, and agricultural activity in Benin, Burkina Faso, Côte d’Ivoire, and Togo. Our main challenge was to obtain welfare and agricultural data that are comparable over time and across countries. Given the low frequency of household consumption surveys and changes in instruments and methodologies between surveys, we can only observe a static pattern of economic geography in the sub- region. In making comparisons across countries, we were limited by a small subset of variables available in all four countries as well as in most administrative units within each country. This clearly narrowed the scope of our analysis. For example, only maize yields could be used as a proxy for agricultural productivity, and farm inputs were limited to fertilizer, land tenure, and land size. As pointed out in Chapter 2, another important caveat concerns the risk of imprecise estimates and lack of representativeness when presenting the data at lower administrative unit levels than the surveys were designed for. Finally, our calculation of a market access index may underestimate coastal areas as well as locations along a country’s borders. On the positive side, efforts are being made to address some of the data-related problems listed above. One of such projects is the West Africa Survey Harmonization Project, which supports the eight member states of WAEMU, including Benin, Burkina Faso, Côte d’Ivoire, and Togo, and conducts harmonized and comparable household surveys using regional standards. The outcomes of this project will help policy makers monitor progress in poverty reduction as well as improvements in agricultural productivity over time and promote regional economic integration among member countries. To deepen our understanding of the economic geography of the sub-region and to hold precise and relevant policy discussions at country levels, further research is needed in the following areas: (a) Geographic poverty traps: A crucial question for any policy designed to tackle poverty is why poor locations stay poor over time, i.e. why poverty traps persist. The concept of poverty traps can be understood as a set of self-reinforcing mechanisms whereby a location starts out poor and remains poor. In other words, current poverty is itself a direct cause of poverty in the future (Azariadis and Stachurski, 2005). While the literature proposes various explanations such as restrictions on labor mobility (Jalan and Ravallion, 2002) and limited availability of production technologies that can lead to higher-income outcomes in poor areas (Kraay and McKenzie, 2014), it is important to study the specific mechanisms of poverty traps in each of the four countries in order to propose meaningful interventions. The outcomes from the West Africa Survey Harmonization Project have the potential to make this research feasible. (b) Labor mobility and migration: We have highlighted the importance of human mobility in economic development and poverty reduction. Like other reallocation mechanisms, labor mobility allows workers to migrate to geographic locations or economic sectors where returns are higher, thus boosting productivity and economic growth. Labor mobility can also empower traditionally disadvantaged groups, especially women (World Bank, 2010). However, there is little information on migration in the four countries or on its role in economic development. A detailed study of labor mobility will help policy makers answer important questions, including the following: (i) In terms of labor mobility across space, how much and how fast have people been migrating from rural to urban areas or from one country to another? What are the implications of any such developments on welfare in the host locations? Is the infrastructure ready and adequate to absorb future inflows of people given current speed? A related study of migration could also focus on constraints on greater and more accelerated agglomerations, including secondary cities, and aim to inform investment projects in urban areas. 70 (ii) In terms of labor mobility across sectors, what patterns and trends are observable as regards structural change in each country? Has there been a transition of labor out of agriculture with the services or manufacturing sectors absorbing this labor? Within each sector, has there been a transition from low-productivity to high-productivity jobs? Most importantly, what has been the role of structural transformation in poverty reduction? (c) Agricultural productivity: Boosting agricultural productivity is a policy priority for alleviating poverty and reducing intra-regional income gaps. Our current analysis focuses on agro-ecological characteristics, agglomeration, market access, and three farm inputs (spending on fertilizer, land tenure, and land size). However, in-depth research is needed to fully understand the determinants of agricultural productivity. Such a study should investigate various measures of agricultural productivity (yields and production value) of all major crops and explore an extensive set of farm inputs such as labor costs, use of irrigation, improved seed, etc. Moreover, such an analysis should be carried out at country level so as to lead to relevant country-specific policy interventions. For example, the classification of agro-ecological or livelihood zones can go beyond the four common zones proposed in this report, thus allowing for detailed policy recommendations on zone-specific input mixes or R&D. (d) Climate change and conflict: Political instability and environmental changes can hamper not only welfare in the affected areas but also in the country as a whole. These issues are especially pronounced in West Africa (Marc, Verjee, and Mogaka, 2015). A complement analysis to this report should overlay maps of conflicts and climate changes on maps of poverty, poverty density, and poverty mass. Such an analysis should aim to assess the roles of political instability and environmental changes in spatial inequality in terms of welfare. If time series data are available, it would be critical to understand how these roles have evolved over time. Moreover, understanding how many poor people have remained trapped in poverty in the affected areas over time as well as how poverty density in these locations has changed can result in relevant policy interventions. Related to the labor mobility study proposed above, the study should focus on a subgroup of migrants, namely refugees, identify who they are in terms of demographics, education, and occupation, and report on the implications of refugee inflows for the economy. (e) Market accessibility: Market access and its relationship to poverty are a determinant factor for investment projects in domains such as infrastructure and public services. For example, an area with a high concentration of poor people with low market access may suggest a positive return from a road construction project. For this purpose, we need to improve our measures, including the market access index introduced in this report. In particular, such an index should take into account access to all transportation modes including coastal ports, airports, roads, railways, and waterways as well as access to markets across borders. In addition, the model evaluating the impact of market accessibility to welfare should aim to address endogeneity that may include better-off households choosing to live in locations with high market accessibility as well as local government in poor locations that cannot afford large investments in infrastructure, thus suffering from low market access. (f) Economic potential across space: This report provides stylized facts concerning spatial differences in welfare. Another innovative and relevant aspect that helps understand the spatial landscape of a country would be a study of the potential for rapid economic development. This study would complement this report by showing the differences between economic potential and performance across the territory. It would be useful for policy makers to have insights into the geographic distribution of levels of economic potential across a given country, the relative strengths and weaknesses of different locations, and the extent to which different locations are fulfilling their potential. The Economic Potential Index (Roberts, 2016) may be one approach for taking this analysis forward. This index captures the extent to which a location possesses five factors that have the potential to contribute to high levels of productivity. These five factors (market access, 71 economic density, urbanization, skills, and local transport connectivity) represent key proximate determinants of local levels of productivity. (g) Regional poverty analysis including Ghana: With its strategic location in the sub-region (i.e. bordered by Côte d’Ivoire, Togo, and Burkina Faso, being located on the same latitude and sharing similar agro-ecological zones as Côte d’Ivoire, Togo, and Benin), and thanks to its membership in the Economic Community of West African States (ECOWAS), Ghana plays an important role in the sub-regional economy. Thus, a regional poverty analysis could be enriched if it were to go beyond AFCF2 countries and include Ghana. 72 References Abro, Zewdu Ayalew, Bamlaki Alamirew Alemu, and Munir Hanjra. 2014. “Policies for Agricultural Productivity Growth and Poverty Reduction in Rural Ethiopia.” World Development 59: 461– 474. Acemoglu, Daron, Simon Johnson, and James A. Robinson. 2001. “The Colonial Origins of Comparative Development: An Empirical Investigation.” American Economic Review 91 (5). Adampoulos, Tasso, and Diego Restuccia. 2017. “Geography and Agricultural Productivity: Cross- Country Evidence from Micro Plot-Level Data.” Working Paper. https://www.economics.utoronto.ca/diegor/research/GAEZ_paper.pdf. AGRHYMET. 2016. “Côte d’Ivoire: Zones et Descriptions des Moyens d’Existence.” CILSS. Akossou, Arcadius, Eloi Attakpa, Noel Fonton, Brice Sinsin, and Roel Bosma. 2016. “Spatial and Temporal Analysis of Maize Crop Yields in Benin from 1987 to 2007.” Agricultural and Forest Meteorology 220 (2016): 177–189. Azariadis, Costas, and John Stachurski. 2005. “Poverty Traps”. Chapter 5. In Handbook of Economic Growth, Vol. 1, Part A, edited by Philippe Aghion and Steven N. Durlauf. Elsevier. Baldwin, Richard, Rikard Forslid, Philippe Martin, Gianmarco Ottaviano, and Frédéric Robert-Nicoud. 2003. Economic Geography and Public Policy. Princeton, NJ: Princeton University Press. Ballon, Paola, Brian Blankespoor, Yeon Soo Kim, and Nobuo Yoshida. n.d. “A New Approach for Evaluating the Impact of Transportation Investment on Market Accessibility and Poverty.” Policy Research Working Paper, World Bank, Washington, DC. Banerjee, A., and E. Duflo. 2007. “The Economic Lives of the Poor”. Journal of Economic Perspectives 21 (1): 141–67 Berg, Claudia N., Brian Blankespoor, and Harris Selod. 2016. “Roads and Rural Development in Sub- Saharan Africa.” Policy Research Working Paper, World Bank, Washington, DC. Binswanger, Hans P., and John McIntire. 1987. “Behavioral and Material Determinants of Production Relations in Land-abundant Tropical Agriculture.” Economic Development and Cultural Change 36 (1): 73–99. Bloom, David E., David Canning, and Jaypee Sevilla. 2003. “Geography and Poverty Traps.” Journal of Economic Growth 8 (4): 355–378. Bosker, Maarten, and Harry Garretsen. 2012. “Economic Geography and Economic Development in Sub- Saharan Africa.” The World Bank Economic Review 26 (3): 443–485. Brakman, S., H. Garretsen, and Ch. Van Marrewijk. 2009. The New Introduction to Geographical Economics. Cambridge: Cambridge University Press. Brinkhoff, Thomas. 2016. citypopulation.de. Buys, Piet, Kenneth M. Chomitz, and Timothy S. Thomas. 2005. “Quantifying the Rural-Urban Gradient in Latin America and the Caribbean.” Research Working Paper 3634, World Bank, Washington, DC. Christiaensen, Luc, Lionel Demery, and Jesper Kuhl. 2011. “The (Evolving) Role of Agriculture in Poverty Reduction: An Empirical Perspective.” Journal of Development Economics 96: 239–354. Christiaensen, Luc, and Gabriel Lawin. 2017. “Maximizing Agriculture's Contribution to the Jobs Agenda.” In Toward Better Employment and Productive Inclusion: A Jobs Diagnostic for Côte 73 d’Ivoire, edited by Luc Christiaensen, and Patrick Premand, 68–84. Washington, DC: World Bank. Combes, Pierre-Philippe, Thierry Mayer, and Jacques-François Thisse. 2008. Economic Geography: The Integration of Regions and Nations. Princeton, NJ: Princeton University Press. Datt, Gaurav, and Martin Ravallion. 1998. “Farm Productivity and Rural Poverty in India.” Journal of Development Studies 34 (4): 62–85. De Brauw, A., V. Mueller, and H. L. Lee. 2014. “The Role of Rural-Urban Migration in the Structural Transformation of Sub-Saharan Africa.” World Development 63: 33–42. De Janvry, Alain, and Elisabeth Sadoulet. 2010. “Agricultural Growth and Poverty Reduction: Additional Evidence.” World Bank Research Observer, 25 (1). Deichmann, Uwe. 1997. Accessibility Indicators in GIS. New York: United Nations. DeLorme. 2015. “Digital Atlas of the Earth 2015.” Diamond, Jared. 1997. Guns, Germs, and Steel. New York: Norton. Dixon, Sam, and Julius Holt. 2010. Livelihood Zoning and Profiling Report: Burkina Faso. Famine Early Warning Systems Networks Report, USAID, Washington, DC. Donaldson, D. Forthcoming. Railroads and the Raj: Estimating the Impact of Transportation Infrastructure. American Economic Review. Donaldson, Dave, and Richard Hornbeck. 2016. “Railroads and American Economic Growth: A Market Access Approach.” Working Paper 19213, National Bureau of Economic Research, Washington, DC. Easterly, William, and Ross Levine. 2002. “Tropics, Germs and Crops: How Endowments Influence Economic Development.” NBER Working Paper 9106. Emran, M. Shahe, and Forhad Shilpi. 2012. “The Extent of the Market and Stages of Agricultural Specialization.” Canadian Journal of Economics 45 (3): 1125–1153. Falkinger, Josef, and Josef Zweimuller. 1996. “The Cross-Country Engel Curve for Product Diversification.” Structural Change and Economic Dynamics 7 (1): 79–97. FAO (Food and Agriculture Organization). 2001. “Country Pasture/Forage Resource Profiles: Burkina Faso.” http://www.fao.org/ag/agp/agpc/doc/Counprof/burkinaFeng.htm. _______. 2009a. “Country Pasture/Forage Resource Profiles: Benin.” http://www.fao.org/ag/agp/agpc/doc/counprof/benin/Benin.htm#3climate. _______. 2009b. “Country Pasture/Forage Resource Profiles: Côte d’Ivoire.” http://www.fao.org/ag/agp/agpc/doc/counprof/ivorycoast/ivorycoast.htm#climate. FEWSNET (Famine Early Warning Systems Network). 2016. Fuglie, Keith, and Nicholas Rada. 2012. “Constraints to Raising Agricultural Productivity in Sub-Saharan Africa.” In Productivity Growth in Agriculture: An International Perspective, edited by K. Fuglie, S. Wang, and V. Eldon Ball. Oxfordshire, UK: CAB International. Fujita, Masahisa, Paul R. Krugman, and Anthony J. Venables. 1999. The Spatial Economy: Cities, Regions, and International Trade. Cambridge, MA: MIT Press. Fujita, Masahisa, and Jacques-François Thisse. 2002. Economics of Agglomeration. Cambridge: Cambridge University Press. 74 Funk, Chris, Pete Peterson, Martin Landsfeld, Diego Pedreros, James Verdin, Shraddhanand Shukla, Gregory Husak, James Rowland, Laura Harrison, Andrew Hoell, and Joel Michaelsen. 2015. “The Climate Hazards Infrared Precipitation with Stations – A New Environmental Record for Monitoring Extremes.” Scientific Data 2, article number 150066. Gallup, John Luke, and Jeffrey D. Sachs. 2001. “The Economic Burden of Malaria” supplement, American Journal of Tropical Medicine & Hygiene 64 (1–2). Gallup, John Luke, Jeffrey D. Sachs, and Andrew D. Mellinger. 1999. “Geography and Economic Development.” International Regional Science Review 22 (2): 179–232. Gavin, Michael, and Ricardo Hausmann. 1998. “Nature, Development, and Distribution in Latin America: Evidence on the Role of Geography, Climate, and Natural Resources.” Inter-American Development Bank Working Paper 378, Washington, DC. Goldstein, Markus, Kenneth Houngbedji, Florence Kondylis, Michael O’Sullivan, and Harris Selod. 2015. “Formalizing Rural Land Rights in West Africa: Early Evidence from a Randomized Impact Evaluation in Benin.” Policy Research Working Paper 7435, World Bank, Washington, DC. Gollin, Douglas, David Lagakos, and Michael E. Waugh. 2014. “The Agricultural Productivity Gap.” Quarterly Journal of Economics 129 (2): 939–993. Gollin, Douglas, Martina Kirchberger, and David Lagakos. 2016. “Living Standards across Space: Evidence from Sub-Saharan Africa.” Working Paper, World Bank, Washington, DC. Golub, Stephen. 2012. “Entrepot Trade and Smuggling in West Africa: Benin, Togo, and Nigeria.” The World Economy 35 (9): 1139–1161. Govereh, J., and Jayne, T. 2003. “Cash Cropping and Food Crop Productivity: Synergies or Trade-offs?” Agricultural Economics 28 (1): 39–50. doi:10.1111/j.1574-0862.2003.tb00133.x. Govereh, J., T. Jayne, and J. Nyoro. 1999. “Smallholder Commercialization, Interlinked Markets and Food Crop Productivity: Cross-country Evidence in Eastern and Southern Africa.” http://aec3.aec.msu.edu/fs2/ag_transformation/atw_govereh.pdf. Grosh, Margaret, Carlo del Ninno, Emil Tesliuc, and Azedine Ouerghi. 2008. For Protection and Promotion: The Design and Implementation of Effective Safety Nets. World Bank, Washington, DC. Hanson, Gordon. 2005. “Market Potential, Increasing Returns, and Geographic Concentration.” Journal of International Economics 67: 1–24. Harris, Chauncy D. 1954. “The Market as a Factor in the Localization of Industry in the United States.” Annals of the Association of American Geographers 44: 315–348. HarvestChoice, International Food Policy Research Institute (IFPRI), and University of Minnesota. 2016. “CELL5M: A Multidisciplinary Geospatial Database for Africa South of the Sahara.” Harvard Dataverse, V2. http://dx.doi.org/10.7910/DVN/G4TBLF. Hausmann, Ricardo. 2001. “Prisoners of Geography.” Foreign Policy 122, 44–53. Henderson, Vernon. 2014. “Urbanization and the Geography of Development.” Policy Research Working Paper 6877, World Bank, Washington, DC. Hijmans, R. J., S. E. Cameron, J. L. Parra, P.G. Jones, and A. Jarvis. 2005. “Very High Resolution Interpolated Climate Surfaces for Global Land Areas.” International Journal of Climatology 25: 1965–1978. 75 Irz, Xavier, Lin, Colin Thirtle, and Steve Wiggins. 2001. “Agricultural Productivity Growth and Poverty Alleviation.” Development Policy Review 19 (4): 449–466. Jalan, Jyotsna, and Martin Ravallion. 2004. “Household Income Dynamics in Rural China.” In Insurance Against Poverty, edited by Stefan Dercon, 108–124. Oxford: Oxford University Press. Jekanowski, Mark D., and James K. Binkley. 2000. “Food Purchase Diversity across U.S. Market.” Agribusiness: An International Journal 16 (4): 417–433. Kraay, Aart, and David McKenzie. 2014. “Do Poverty Traps Exist? Assessing the Evidence.” Journal of Economic Perspectives 28 (3): 127–148. Krugman, Paul. 1991. “Increasing Returns and Economic Geography.” Journal of Political Economy 99 (3): 483–99. ———. 1996. The Self-Organizing Economy. Oxford: Blackwell. Krugman, Paul R., and Anthony J. Venables. 1995. “Globalization and the Inequality of Nations.” The Quarterly Journal of Economics 110: 857–880. Lall, Somik V., J. Vernon Henderson, and Anthony J. Venables. 2017. Africa’s Cities: Opening Doors to the World. World Bank, Washington, DC. Lall, Somik V., Zmarak Shalizi, and Uwe Deichmann. 2004. “Agglomeration Economies and Productivity in Indian Industry.” Journal of Development Economics 73 (2004): 643–73. Lee, Jonq-Ying, and Mark G. Brown. 1989. “Consumer Demand for Food Diversity.” Southern Journal of Agricultural Economics 1989: 47–53. Lichter, Daniel T., Domenico Parisi, and Michael C. Taquino. 2012. “The Geography of Exclusion: Race, Segregation, and Concentrated Poverty.” Social Problems 59 (3): 364–388. Ligon, Ethan, and Elisabeth Sadoulet. 2007. “Estimating the Effects of Aggregate Agricultural Growth on the Distribution of Expenditures.” Background Paper. WPR 2008. Limao, Nuno, and Anthony J. Venables. 2001. “Infrastructure, Geographical Disadvantage, Transport Costs, and Trade”. The World Bank Economic Review 15 (3): 451–479. Lusk, Jayson L., Jutta Roosen, and Jason E. Shogren. 2011. The Oxford Handbook of the Economics of Food Consumption and Policy. Oxford University Press. Marc, Alexandre, Neelam Verjee, and Stephen Mogaka. 2015. The Challenge of Stability and Security in West Africa. Washington, DC: World Bank and Agence Française de Développement. https://openknowledge.worldbank.org/handle/10986/22033. Marshall, Alfred. 1920. Principles of Economics. London: MacMillan. Mayer, Thierry. 2008. “Market Potential and Development.” Working Paper DP 6798, Center for Economic Policy Research, Washington, DC. McMillan, M. S., and K. Harttgen. 2014. “What is Driving the ‘African Growth Miracle?” Working Paper 20077, National Bureau of Economic Research, Washington, DC. Mellinger, Andrew D., Jeffrey D. Sachs, and John L. Gallup. 2000. “Climate, Coastal Proximity, and Development.” In Oxford Handbook of Economic Geography, edited by Gordon L. Clark, Maryann P. Feldman, and Meric S. Gertler. Oxford University Press. Ministère de l’Environnement et des Ressources Forestières. 2003. “Stratégie de Conservation et d’Utilisation Durables de la Diversité Biologique.” Lomé. _______. 2014. “Cinquième Rapport National sur la Diversité Biologique du Togo 2009–2014.” Lomé. 76 Minten, Bart, and Christopher Barrett. 2008. “Agricultural Technology, Productivity, and Poverty in Madagascar.” World Development 36 (5): 797–822. Minten, B., L. Randrianarison, and J. F. Swinnen. 2009. “Global Retail Chains and Poor Farmers: Evidence from Madagascar.” World Development 37 (11): 1728–1741. doi:10.1016/j.worlddev.2008.08.024. Nacoulma, Jean Didier, and Jean Bruno Guigma. 2015. Institutional Context of Soil Information in Benin. The International Center for Tropical Agriculture (CIAT) Report. OECD (Organisation for Economic Cooperation and Development). 1994. Creating Rural Indicators for Shaping Territorial Policies. Paris: OECD Publications. Partridge, Mark D., and Dan S. Rickman. 2006. The Geography of American Poverty: Is There a Need for a Place-Based Policies? Michigan. Porter, Michael. 1998. “Clusters and the New Economics of Competition.” Harvard Business Review, November–December, 77–90. Puga, Diego. 1999. “The Rise and Fall of Regional Inequalities.” European Economic Review 43: 303– 334. Radelet, Steven C., and Jeffrey D. Sachs. 1998. Shipping Costs, Manufactured Exports, and Economic Growth. Cambridge, MA: Harvard Institute for International Development. Roberts, Mark. 2016. “Identifying the Economic Potential of Indian Markets.” World Bank Policy Research Working Paper 7623. Rodrik, Dani, Arvind Subramanian, and Francesco Trebbi. 2004. “Institutions Rule: The Primacy of Institutions over Geography and Integration in Economic Development.” Journal of Economic Growth 9 (2): 131–165. Rosenthal, Stuart S., and William C. Strange. 2004. “Chapter 49 Evidence on the Nature and Sources of Agglomeration Economies.” In Handbook of Regional and Urban Economics, Vol. 4, 2119–71. Amsterdam: Elsevier. Sachs, Jeffrey D. 2000 “Tropical Underdevelopment.” Prepared for Economic History Association Annual Meeting. CID Working Paper 57. http://www.cid.harvard.edu/cidwp/057.htm. Sachs, Jeffrey D., and Pia Malaney. 2002. “The Economic and Social Burden of Malaria.” Nature Insight 415 (6872). Sachs, Jeffrey D., Andrew D. Mellinger, and John L. Gallup. 2001. “The Geography of Poverty and Wealth.” Scientific American, 71–74. Schneider, Kate, and Mary Kay Gugerty. 2010. “The Impact of Export-Driven Cash Crops on Smallholder Households.” Evans School Policy Analysis Research Brief 94. ———. 2011. “Agricultural Productivity and Poverty Reduction: Linkages and Pathways.” Evans School Review 1 (1): 56–74. Smith, Adam. 1776. The Wealth of Nations. New York: Simon & Brown. Strasberg, Paul J., T. Jayne, Takashi Yamano, James Nyoro, Daniel Karanja, and John Strauss. 1999. “Effects of Agricultural Commercialization on Food Crop Input Use and Productivity in Kenya.” MSU International Development Working Paper 71. Thériault, Veronique, Alpha Kergna, Abramane Traore, Bino Teme, and Melinda Smale. 2015. “Review of the Structure and Performance of the Fertilizer Value Chain in Mali.” Food Security Policy Innovation Lab Working Paper Mali-2015-2, Michigan State University, East Lansing, MI. 77 Uchida, Hirotsugu, and Andrew Nelson. 2010. “Agglomeration Index: Toward a New Measure of Urban Concentration.” Working Paper 2010/29, United Nations University, Tokyo. United Nations. 2014. “World Urbanization Prospect, the 2014 Revision.” United Nations, Department of Economic and Social Affairs, Population Divisions, New York. https://esa.un.org/unpd/wup/DataSources. van de Walle, Dominique. 2013. “Lasting Welfare Effects of Widowhood in Mali.” World Development 51: 1–19. ———. 2015. “Women Left Behind? Poverty and Headship in Africa.” Working Paper 7331, World Bank, Washington, DC. Venables, Anthony J. 1996. “Equilibrium Locations of Vertically Linked Industries.” International Economic Review 37: 341–59. Vissoh, Pierre V., Gualbert Gbéhounou, Adam Ahanchede, Thom W. Kuyper, and Niels G. Röling. 2004. “Weeds as Agricultural Constraint to Farmers in Benin: Results of a Diagnostic Study.” Wagningen Journal of Life Science 52 (3/4): 305–329. World Bank. 2008. Agriculture for Development. World Development Report. Washington, DC: World Bank. ———. 2009. Reshaping Economic Geography. World Development Report. Washington, DC: World Bank. ———. 2010. The Poor Half Billion in South Asia – What is Holding Back Lagging Regions? Washington, DC: World Bank. ———. 2017. World Development Indicators. Washington, DC: World Bank. World Food Programme. 2014. “E-voucher for Food Security – A Potential for India’s Social Safety Nets?” World Food Programme. Yoshida, Nobou, and Uwe Deichmann. 2009. “Measurement of Accessibility and Its Application.” Journal of Infrastructure Development 1 (1): 1–16. 78 Appendix A: Agro-ecological Zones Map A.1: Benin Source: Data from FAO (2009a), image from Akossou et al. (2016) 79 Map A.2: Burkina Faso Source: FEWSNET (2016). Map A.3: Togo Source: Ministère de l’Environnement et des Ressources Forestières (201 4). 80 Map A.4: Côte d’Ivoire Source: AGRHYMET (2016) 81 Appendix B: Market Accessibility Index – Methodology Measuring Access to Regional Markets Following the standard approach to calculating access to markets in the literature, the domestic market access for a given location along the road network is a function of the weighted sum of populated places in all other locations discounted by travel time on the road.20 Formally, we define market access in a location i ( ): = ∑ − (1) where is the population in location j, is travel time between locations i and j, and is a trade elasticity parameter. Following Donaldson (2010), we use elasticity of trade, , equal to 3.8 for equation (1) of the classical model. Following other regional work with geographically limited populated places data (e.g., Lall, Shalizi, and Deichmann, 2004; Yoshida et al., 2009, Ballon et al., n.d.), we define another market access according to the Negative Exponential model (see Deichmann 1997) in location i ( ): − − ( ⁄ ) = ∑ . 2 2 (2) where is the population in location j, is travel time between locations i and j, and a and b are trade elasticity parameters based on Deichmann (1997). We use the negative exponential model in equation (2) and employ the parameters of a = 20 and 30, and b = 2. We then summarize market access at an administrative level for each country by transforming the market access results from both functional forms to an inverse distance weighted grid and taking the mean of the grid in the administrative level for each model. Data The administrative boundary data are taken from the statistical services of each country or from a global database in order to match the boundaries with the aggregated census data. The boundary files of 37 prefectures in Togo are taken from the National Statistical Agency. The 2015 boundary document for 108 departments in Côte d’Ivoire is taken from the National Institute for Statistics. The boundary document for 78 communes in Benin is taken from FAO.21 The 2006 boundary document for 45 provinces in Burkina Faso is taken from the National Institute for Statistics and Demography. We used geo-referenced city-level estimated population data from the set compiled by Blankespoor, Khan, and Selod (n.d.). The primary source for their city population levels data is the census estimates compiled by Brinkhoff (2016) and subsequently geo-referenced in order to add the spatial dimension. These data provide estimates for 2015 by using a constant continuous growth rate for cities derived from inter-censal population data. We impose a minimum population of greater than or equal to 10,000, which yields 206 cities in the four countries under study as locations of regional markets.22 The sum of population from these locations captures between 27 and 46 percent of the total country-level population from World Bank World Development Indicators (WDI) (Table B.), which suggests that the rural population makes up a large share of the total and that the available data do not sufficiently account for local markets. We modify the 20 Examples from the literature with similar market access include Harris (1954), Hanson (2005), Emran and Shilpi (2012), Jedwab and Storeygard (2015), Blankespoor et al. (2016), Berg, Blankespoor, and Selod (2016), and Donaldson and Hornbeck (2016). 21 We merge the urban and rural boundaries for Djougou commune in Benin in order to harmonize the data with development indicators. 22 We exclude five cities that lack inter-censal population data on the citypop website. 82 coordinates slightly so that each populated location corresponds to a node on the road network in order to calculate market access in Formula (1) (see above). Table B.1: Population in city database for each country Share of urban Country 2015 Population (WDI) population (%) Benin 10,879,829 0.357 Burkina Faso 18,105,570 0.278 Côte d’Ivoire 22,701,556 0.460 Togo 7,304,578 0.351 The roads data we used are from DeLorme (2015), which provides a regionally consistent and well- connected geometry of road segments for network analysis. The functional road categories in the road data include primary, secondary, and tertiary roads. Derived from estimates in Jedwab and Storeygard (2016), we assume the following speeds by road categories: 60 km/h for primary roads, 40 km/h for secondary roads,23 12 km/h for tertiary roads, and 5 km/h as background travel speed in the absence of a road.24 Finally, we make a number of modifications to the input data. First, we combine the spatial distribution of the cities and road segments by using a near function to force the city locations to coincide with the nearest road node. Second, we construct additional nodes from the intersection of a 5-kilometer radius from each city and the radial roads in order to ensure a wider geographic coverage of model results. Results: Access to Regional Markets Both models produce a skewed distribution of market access across the four countries. The capital and large cities have primary regional radial roads, whereas the majority of administrative regions do not have strong connections to these regional markets (proxied by the set of approximately 200 cities). Following recent literature using the classical model (Jedwab and Storeygard, 2015; Berg, Blankespoor, and Selod, 2016; Blankespoor et al. 2016), a robustness check with alternative values (e.g., 8.2) of the trade elasticity provides similar results to theta 3.8.25 Following the negative exponential model, a robustness check provides similar results for a = 20 and a = 30. When comparing the two models, Burkina Faso and Côte d'Ivoire have a strong correlation to themselves compared to the results from Benin and Togo.26 23 Since DeLorme (2015) does not include the surface type of the road, we use 50km/h as the average of Jedwab and Storeygard’s (2016) estimates for paved road and improved road. 24 We also consulted and made minor modifications based on the road data available from the Africa Infrastructure Country Diagnostic Study (Foster and Briceño-Garmendia, 2010). 25 The model results with theta 3.8 or 8.2 have a correlation of 0.98 or greater by country. 26 The correlations between the two models are: BFA 0.94, CIV 0.95, BEN 0.73, and TGO 0.74. 83 Appendix C: Extra Materials Figure C.1: Poverty rates by administrative unit 84 85 86 Figure C.2: Poverty density by administrative unit 87 88 89 Figure C.3: Key crops across zones Proportion of individuals planting a crop across zones 90 Table C.1: Proportion of individuals growing crops across zones (top 5 crops) Côte d’Ivoire Burkina Faso Togo Benin Proportion Proportion Crop Proportion Proportion Zone Crop Crop Crop (%) (%) (%) (%) 1 Yam 67 Mil 76 Maize 89 Cereals 95 1 Cashew 45 Sorghum 70 Sorghum 59 Cotton 68 1 Cassava 18 Cowpea 33 Beans, Cowpea 53 Tubers 53 1 Peanut 15 Peanut 28 Okra 50 Fruits and 47 vegetables 1 Cocoa 14 Sesame 20 Rice 46 Palm oil 5 2 Maize 56 Sorghum 88 Maize 88 Cereals 92 2 Fluvial 49 Cowpea 72 Beans, Cowpea 45 Tubers 83 rice 2 Cotton 49 Mil 63 Sorghum 44 Cotton 43 2 Cashew 42 Peanut 60 Yam 43 Fruits and 42 vegetables 2 Peanut 39 Maize 33 Soya 34 Palm oil 2 3 Cocoa 73 Maize 72 Maize 93 Cereals 91 3 Yam 24 Sorghum 71 Beans, Cowpea 66 Tubers 85 3 Rice (bas 23 Peanut 39 Cassava 42 Fruits and 61 fond) vegetables 3 Cassava 21 Mil 34 Yam 39 Palm oil 16 3 Coffee 20 Cowpea 30 Sorghum 22 Cotton 15 4 Cocoa 72 Maize 83 Maize 94 Cereals 91 4 Yam 18 Sorghum 58 Cassava 45 Tubers 76 4 Rice (bas 18 Cotton 38% Palm nuts 26 Fruits and 60 fond) vegetables 4 Cassava 16 Peanut 36% Peanut 23 Palm oil 49 4 Coffee 13 Sesame 33% Beans, Cowpea 22 Cotton 3 91 Table C.2: Poverty and other indicators, by zone, with standard errors Benin Burkina Faso Cote d'Ivoire Togo Zone 1 Zone 2 Zone 3 Zone 4 Zone 1 Zone 2 Zone 3 Zone 4 Zone 1 Zone 2 Zone 3 Zone 4 Zone 1 Zone 2 Zone 3 Zone 4 poverty rates 0.776 0.831 0.659 0.619 0.354 0.463 0.424 0.429 0.302 0.442 0.335 0.212 0.746 0.598 0.603 0.237 (0.012) (0.007) (0.008) (0.005) (0.015) (0.007) (0.013) (0.009) (0.009) (0.010) (0.010) (0.006) (0.023) (0.026) (0.021) (0.013) Access to services cellphone 0.733 0.719 0.790 0.825 0.804 0.863 0.762 0.749 0.664 0.699 0.654 0.814 0.576 0.751 0.659 0.640 (0.015) (0.009) (0.009) (0.005) (0.023) (0.008) (0.020) (0.014) (0.020) (0.016) (0.021) (0.013) (0.033) (0.037) (0.031) (0.036) electricity 0.136 0.143 0.285 0.403 0.014 0.196 0.052 0.169 0.518 0.468 0.396 0.771 0.123 0.443 0.272 0.733 (0.009) (0.006) (0.008) (0.005) (0.004) (0.006) (0.006) (0.007) (0.009) (0.010) (0.011) (0.006) (0.017) (0.027) (0.019) (0.013) improved toilet 0.136 0.139 0.284 0.511 0.154 0.573 0.187 0.565 0.631 0.625 0.688 0.882 0.167 0.432 0.248 0.835 (0.009) (0.006) (0.007) (0.005) (0.011) (0.007) (0.010) (0.009) (0.009) (0.010) (0.010) (0.004) (0.019) (0.027) (0.019) (0.011) piped water 0.214 0.194 0.248 0.411 0.014 0.175 0.016 0.092 0.329 0.113 0.118 0.510 0.065 0.323 0.248 0.399 (0.011) (0.007) (0.007) (0.005) (0.004) (0.006) (0.003) (0.005) (0.009) (0.006) (0.007) (0.007) (0.013) (0.025) (0.019) (0.015) Share of employed working-aged individual in each sector agriculture 0.716 0.773 0.641 0.386 0.951 0.728 0.921 0.796 0.535 0.565 0.589 0.378 0.491 0.385 0.628 0.097 (0.013) (0.008) (0.008) (0.005) (0.007) (0.007) (0.007) (0.008) (0.010) (0.010) (0.011) (0.007) (0.040) (0.030) (0.030) (0.011) industry 0.065 0.066 0.118 0.226 0.017 0.069 0.014 0.060 0.122 0.197 0.137 0.191 0.075 0.159 0.133 0.264 (0.007) (0.005) (0.006) (0.004) (0.004) (0.004) (0.003) (0.005) (0.006) (0.008) (0.008) (0.006) (0.021) (0.023) (0.021) (0.016) service 0.166 0.119 0.149 0.273 0.032 0.202 0.065 0.143 0.338 0.236 0.270 0.424 0.311 0.354 0.188 0.469 (0.011) (0.006) (0.006) (0.005) (0.006) (0.006) (0.007) (0.007) (0.009) (0.009) (0.010) (0.007) (0.037) (0.030) (0.025) (0.018) other sector 0.052 0.042 0.093 0.116 0.000 0.000 0.000 0.000 0.005 0.002 0.004 0.007 0.122 0.102 0.050 0.170 (0.007) (0.004) (0.005) (0.003) 0.000 0.000 0.000 0.000 (0.001) (0.001) (0.001) (0.001) (0.026) (0.019) (0.014) (0.014) Housing conditions house ownership 0.939 0.952 0.874 0.880 0.608 0.784 0.690 0.586 0.496 0.626 0.571 0.238 (0.008) (0.004) (0.007) (0.005) (0.020) (0.014) (0.021) (0.017) (0.034) (0.042) (0.032) (0.032) concrete roof 0.006 0.008 0.010 0.028 0.000 0.003 0.003 0.002 0.011 0.002 0.007 0.007 0.003 0.004 0.001 0.138 (0.003) (0.002) (0.002) (0.002) (0.001) (0.001) (0.002) (0.002) (0.004) (0.002) (0.004) (0.003) (0.004) (0.006) (0.002) (0.026) concrete wall 0.191 0.395 0.477 0.449 0.024 0.054 0.061 0.116 0.588 0.544 0.392 0.500 0.058 0.186 0.213 0.617 (0.014) (0.010) (0.011) (0.007) (0.009) (0.006) (0.011) (0.010) (0.021) (0.017) (0.022) (0.017) (0.016) (0.034) (0.027) (0.037) Household demographics household size 7.453 7.432 6.662 5.629 12.508 14.290 10.847 11.423 7.043 8.227 5.501 6.558 8.461 7.421 7.162 6.825 (0.130) (0.069) (0.070) (0.037) (0.333) (0.196) (0.198) (0.210) (0.138) (0.171) (0.108) (0.102) (0.254) (0.289) (0.260) (0.292) share of working-age 0.440 0.431 0.487 0.514 0.443 0.494 0.434 0.489 0.540 0.535 0.589 0.613 0.447 0.511 0.485 0.566 (0.005) (0.003) (0.004) (0.002) (0.005) (0.003) (0.005) (0.003) (0.004) (0.005) (0.005) (0.003) (0.009) (0.012) (0.009) (0.007) dependency ratio 1.780 1.774 1.531 1.488 1.666 1.575 1.850 1.492 1.737 1.505 1.370 1.412 1.630 1.536 1.487 1.416 (0.034) (0.021) (0.022) (0.014) (0.053) (0.021) (0.049) (0.029) (0.053) (0.032) (0.045) (0.035) (0.071) (0.088) (0.059) (0.075) Household head demographics male 0.934 0.926 0.842 0.802 0.958 0.944 0.936 0.912 0.715 0.915 0.832 0.828 0.888 0.757 0.746 0.676 (0.008) (0.005) (0.008) (0.006) (0.012) (0.006) (0.011) (0.009) (0.019) (0.009) (0.017) (0.013) (0.021) (0.037) (0.028) (0.035) married 0.922 0.904 0.878 0.881 0.963 0.921 0.938 0.902 0.662 0.937 0.724 0.771 0.865 0.794 0.896 0.744 (0.009) (0.006) (0.007) (0.005) (0.011) (0.007) (0.011) (0.010) (0.020) (0.008) (0.020) (0.014) (0.023) (0.035) (0.020) (0.033) * standard errors in bracket 92 Appendix D: Summary of Findings on Agricultural Activities across Countries, by Zone Côte d’Ivoire Each zone is distinguished by the cash crops produced: cashew in Zone 1, cotton in Zone 2, and cacao in Zones 3 and 4. Zone 1 is widely disadvantaged, with lower yields, less input use, and lower revenues from sales. Zone 1 Cashew is produced in this zone. This is a largely disadvantaged zone. Maize and yam yields are lowest in this zone, as are input and land use (with less spending on fertilizer, less use of pesticides and irrigation, and smaller farms). The value of sales in this zone is also the lowest. Zone 2 Cotton is produced in this zone. This is a zone with the highest input use and largest farm sizes. Alongside Zone 4, this zone has the highest revenues from crop sales. Zones 3 & 4 Cacao is produced in both of these zones, and rice production is also prevalent. Both zones have very high maize yields. Zone 4 has relatively higher revenues from crop sales than Zone 3. Compared to Zones 1 and 2, there is stronger securitization of land in Zones 3 and 4. Burkina Faso Sorghum is prevalent across all zones, and cotton is prevalent in Zone 4, the most advantaged zone. Zone 1 is widely disadvantaged, but has better maize yields than Zones 2 and 3 and better securitization of land than all other zones. Zone 1 This zone has the lowest input use (less spending on fertilizer and less use of pesticides), mainly subsistence farmers (only 30 percent sell some of their crop), and low revenues from sales. However, maize yields are higher than in Zone 1 and 2, and land securitization is stronger than in all other zones. Zones 2 & 3 Maize yields are lowest in these zones, but input use and output sales are higher compared to Zone 1. There is some production of cotton in Zone 3. Zone 4 Cotton is widely produced here. This zone somewhat mirrors Zone 2 in Côte d’Ivoire. This zone has the highest maize yields and input use, the highest proportion of farmers selling some of their crop, and the highest revenues from crop sales. However, securitization of land is the weakest. Togo There is less variation across regions in the dominant crops, with maize, root crops (yam and cassava), and cowpeas being common throughout the country. Pesticide use is low throughout the country. However, Zone 1 has better prospects for agriculture, whereas Zone 2 seems to be disadvantaged. Zone 1 This zone has rice and soya production, and has the largest value of sales. It also has higher spending on fertilizer (comparable to spending on fertilizer in Zone 2). Zones 2 & 4 These zones have fewer individuals working in agriculture or selling some of their produce. These zones also have lower sales value of produce (but better securitization of land). Zone 2 has higher spending on fertilizer than Zone 4. This zone also has some soya production. Zone 3 This zone has the lowest maize yields as well as low spending on fertilizer. Despite this, the value of sales is higher compared to Zones 2 and 4. Benin The zones are divided across cash crops, with cotton in Zones 1 and 2, palm oil in Zone 4, and to a lesser extent some cotton and palm oil production in Zone 3. In terms of yields, Zones 1 and 2 seem to be more advantaged, whereas Zone 4 less advantaged in terms of yields. Zone 1 Cotton is largely produced in this zone, and with higher yields. Irrigation is slightly higher in Zone 1, but is low throughout the country. Zone 2 Cotton is widely produced in this zone, but with lower yields compared to Zone 1. However, maize yields are highest in this zone. Spending on fertilizer is also highest in this zone. Zone 3 This zone has little cash crop production, but it has some palm oil and cotton production. Zone 4 This zone is prevalent in palm oil production. Maize and cotton yields are the lowest in this zone, as is fertilizer spending. Securitization of land is better in this zone as well as in Zone 3 compared to Zones 1 and 2. 93 Appendix E: Agricultural Data – Notes on Model Construction For Côte d’Ivoire, there are two relevant parts to the survey (particularly how the database is organized): a land section (harvested plots) and a crops section (crops planted on those harvested plots). Although these two sections should in theory represent the same number of households, in practice they do not. The land section coincides with 6,849 households, while the crops section coincides with only 3,402 households. Additional caution is recommended when interpreting data that use the crops section for Côte d’Ivoire (such as yields, sales, and input use, but not irrigation, which comes from the land section). Maize and Cash Crop Yields For each country, yield is a measure of average kilograms per hectare of a crop produced in a given agro- ecological zone; however, due to data constraints, the yield measure is calculated slightly differently for each country. • Côte d’Ivoire: Conversion into kilograms for non-standard output units was not available in the survey. Thus, the conversion factor for a given non-standard unit was calculated as the ratio of the average (national level) sales price for a kilogram of maize and the average (national level) sales price for a given non-standard unit of maize. We chose to calculate a national level conversion factor (as opposed to district level) due to constraints on the number of observations for sales of non-standard units at a sub-national level. However, district-level conversion factors were used for cacao yields in Côte d’Ivoire. • Burkina Faso: Conversion into kilograms for non-standard output units was not available in the survey. Thus, the conversion factor for a given non-standard unit was calculated as the ratio of the average (regional level) sales price for a kilogram of maize and the average (regional level) sales price for a given non-standard unit of maize. Thus, for Burkina Faso, region-crop-specific conversion factors were used when converting non-standard units into kilograms. The same method was used when calculating cotton yields for Burkina Faso. • Togo: Conversion into kilograms for non-standard output units was reported by respondents in the survey. For each non-standard unit, the conversion factor was calculated as the regional median of the reported conversion factors. Thus, region-crop-specific conversion factors were used when converting non-standard units into kilograms. The conversion factors were self-reported as opposed to assumed from sales and price data. Due to data constraints, assumptions had to be made on how much land each household allocated to maize (or any given crop). Since the survey did not elicit this information directly, we made the assumption that for any given plot with more than one crop, the main crop is assigned 70 percent of the plot area and the secondary crop is assigned 30 percent of the plot area. The survey does not contain information on whether a third crop is present in any given plot. • Benin: Yields were not taken from the household survey but from data from the Ministry of Agriculture on commune-level estimates of production and land used for various crops. The key difference in comparability is that the Côte d’Ivoire, Burkina Faso, and Togo databases may reflect smallholder farmers as the data come from the household survey, while the Benin data may reflect larger farms. Caution is recommended when making comparison of maize and cotton yields between Benin and the three other countries in the sub-region. To glean information from Benin on smallholder farmers, we used the PFR 2011 database, which consists of plot-level data from a non-representative household survey from 2011 conducted to study the effects of land titling. 94 Table E.1 Comparison of yield data from Ministry of Agriculture and PFR household survey Ministry of Agriculture data (2015) PFR 2011 data (smallholders) Zone 1 1,280 kg/ha 1,337 kg/ha Zone 2 1,582 kg/ha 1,162 kg/ha Zone 3 1,272 kg/ha 925 kg/ha Zone 4 1,075 kg/ha 1,205 kg/ha Overall, the yields of smallholder farmers in 2011 are close to those at the commune level in 2015, though patterns may be different across zones. Based on 2015 data from the Ministry of Agriculture, Zone 2 had the highest yields and Zone 4 the lowest. Based on the 2011 PFR, Zone 1 had the highest yields and Zone 3 the lowest. However, we cannot draw much from this analysis as the datasets are different in key aspects, including representativeness and time of survey. 95