Mapping Poverty in Sudan August 2019 Poverty and Equity Global Practice Africa Region Standard Disclaimer: This volume is a product of the staff of the International Bank for Reconstruction and Development/The World Bank. The findings, interpretations, and conclusions expressed in this paper do not necessarily reflect the views of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. . Copyright Statement: The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. The International Bank for Reconstruction and Development/ The World Bank encourages dissemination of its work and will normally grant permission to reproduce portions of the work promptly. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA, telephone 978-750- 8400, fax 978-750-4470, http://www.copyright.com/. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World Bank, 1818 H Street NW, Washington, DC 20433, USA, fax 202-522-2422, e-mail pubrights@worldbank.org. Abstract This report aims to map poverty and inequality in Sudan and would be representative of the 18 states and 131 localities of Sudan. The poverty mapping technique is based on a small area estimation (SAE) technique developed by the World Bank to derive estimates of geographic poverty and inequality. It combines data from the 2014/15 National Household Budget and Poverty Survey (NHBPS) and the 2008 Population and Housing Census data to build spatially disaggregated poverty maps. Although household surveys usually include measures of income and wealth, they are not representative beyond the state level. Yet, allowing lower levels of disaggregation is important for policy interventions, particularly for countries like Sudan that have state governments, which manage the activities of the state while reporting to the federal government. This study uses a model of household expenditure from a survey data set to estimate household welfare at the lower levels and apply it to the census data set which does not provide information on household income or expenditure. These maps illustrate the information gains provided by SAE, show there is a substantial spatial heterogeneity within the localities, and highlight the small areas most likely to exhibit the highest risk of poverty. i Abbreviations AfDB African Development Bank AIC Akaike Information Criteria CBS Central Bureau of Statistics EB Empirical Best ELL Model suggested by Elbers, Lanjouw, and Lanjouw FE Fixed-effects FGT Foster-Greer-Thorbecke GDP Gross Domestic Product GLS General Least Squares HCI Human Capital Index HDI Human Development Index MDG Millennium Development Goal MSE Mean Square Error NHBPS National Household Budget and Poverty Survey OLS Ordinary Least Squares PSU Primary Sampling Unit SAE Small Area Estimation VIF Variance Inflation Factor ii This report was prepared by Alvin Etang Ndip (Senior Economist, GPV01), Minh Cong Nguyen (Senior Data Scientist, GPV03), Ando Rahasimbelonirina (Consultant, GPV01), and Tarig Hashim (Geographic Information System [GIS] Specialist, Central Bureau of Statistics [CBS]). Overall guidance was provided by Pierella Paci (Practice Manager, GPV01). The authors would like to thank the CBS for providing very useful feedback on initial drafts. In particular, many thanks to Dr. Karamalla Ali Abdelrahman (Director General, CBS), Somaia Khalid (Director, Methodology Directorate, CBS), Huda Mohamed Osman (Senior Information Technology [IT] Staff, CBS), and Enaam Mubarak (IT Staff, CBS). The authors would also like to thank Nobuo Yoshida (Lead Economist, GPV01) and Rose Mungai (Senior Economist/Statistician, GPV03) for very useful peer reviewer comments. The report also benefited from comments from Eiman Adil Mohamed Osman (Consultant, GPV01) and Fareed Hassan (Consultant, GWA08). Vice President Hafez Ghanem Country Director Carolyn Turk Senior Director Carolina Sanchez-Paramo Practice Manager Pierella Paci Task Team Leaders Alvin Etang Ndip iii Table of Contents Abstract ................................................................................................................................. i 1. Introduction ................................................................................................................ 1 2. Methodology and Data................................................................................................ 3 2.1. Methodology................................................................................................................................. 3 2.2. Main Sources of Data .................................................................................................................... 6 2.2.1 Census and NHBPS ....................................................................................................................... 6 2.2.2 Matching NHBPS and Census Data .............................................................................................. 6 2.3. Modeling for Monetary Poverty ................................................................................................... 8 2.4. Technical Challenges ................................................................................................................... 14 3. Constructing the 2014/15 Sudan Poverty Maps ......................................................... 16 3.1. Model Selection .......................................................................................................................... 16 3.2. Level of Disaggregation ............................................................................................................... 23 4. Poverty Mapping Results ........................................................................................... 24 5. Conclusions ............................................................................................................... 38 References .......................................................................................................................... 39 Appendix A: Sudan Administrative Boundaries .................................................................... 41 Appendix B: Common Variables between the Census and 2014/15 NHBPS ........................... 42 Appendix C: Region Alpha Model Estimates ......................................................................... 48 Appendix D: Poverty Measures ............................................................................................ 51 Appendix E: Census Poverty Measures by Administrative Units ............................................ 53 Appendix F: Census Non-monetary Indicators by Administrative Units ................................. 60 List of Boxes Box 1: Step-by-step Summary of the Modelling Approach .......................................................................... 5 List of Tables Table 1: Geographic Distribution between the Census (2008) and the 2014/15 NHBPS ............................. 9 Table 2: Comparison of Household Characteristics between the Census (2008) and the 2014/15 NHBPS, Weighted Average, Variables Used in Model for Region 1 ........................................................................... 9 Table 3: Comparison of Household Characteristics between the Census (2008) and the 2014/15 NHBPS, Weighted Average, Variables Used in Model for Region 2 ......................................................................... 10 iv Table 4: Comparison of Household Characteristics between the Census (2008) and the 2014/15 NHBPS, Weighted Average, Variables Used in Model for Region 3 ......................................................................... 11 Table 5: Comparison of Household Characteristics between the Census (2008) and the 2014/15 NHBPS, Weighted Average, Variables Used in Model for Region 4 ......................................................................... 12 Table 6: Comparison of Household Characteristics between the Census (2008) and the 2014/15 NHBPS, Weighted Average, Variables Used in Model for Region 5 ......................................................................... 12 Table 7: Comparison of Household Characteristics between the Census (2008) and the 2014/15 NHBPS, Weighted Average, Variables Used in Model for Region 6 ......................................................................... 13 Table 8: Model Estimates Based on the 2014/15 NHBPS - ‘Northern’ (Beta Model, Region 1) ................. 17 Table 9: Model Estimates Based on the 2014/15 NHBPS - ‘Eastern’ (Beta Model, Region 2) .................... 18 Table 10: Model Estimates Based on the 2014/15 NHBPS - ‘Khartoum’ (Beta Model, Region 3) .............. 18 Table 11: Model Estimates Based on the 2014/15 NHBPS - ‘Central’ (Beta Model, Region 4) .................. 19 Table 12: Model Estimates Based on the 2014/15 NHBPS - ‘Kordofan’ (Beta Model, Region 5) ............... 20 Table 13: Model Estimates Based on the 2014/15 NHBPS – ‘Darfur’ (Beta Model, Region 6)................... 21 Table 14: Poverty Estimates from Survey (Observed) and the Census (SAE) ............................................. 24 Table 15: Census SAE of Poverty and Gini at the National and Regional Levels......................................... 24 Table 16: Census SAE of Poverty and Gini at the State Level ..................................................................... 25 Table 17: Census SAE of Poverty and Gini at the Locality Level ................................................................. 30 Table C.1: Northern (Alpha Model, Region 1)............................................................................................. 48 Table C.2: Eastern (Alpha Model, Region 2) ............................................................................................... 48 Table C.3: Khartoum (Alpha model, Region 3)............................................................................................ 49 Table C.4: Central (Alpha model, Region 4) ................................................................................................ 49 Table E.1: Poverty Measures by Region and State ..................................................................................... 53 Table E.2: Poverty Measures by Locality .................................................................................................... 54 Table F.1: Population Characteristics by Region and State ........................................................................ 60 Table F.2: Households Characteristics by Region and State ....................................................................... 61 List of Figures Figure 1: Distributions with Actual and Imputed Testing Sample .............................................................. 22 Figure 2: Weighted Ratio Mean Square Error of Outsample for Sudan and its Regions ............................ 23 List of Maps Map 1: Direct Estimates at the State Level ................................................................................................... 2 Map 2: Census SAE of Poverty at the State Level ....................................................................................... 27 Map 3: Census SAE of Number of Poor at the State Level.......................................................................... 28 Map 4: Census SAE of Gini at the State Level ............................................................................................. 29 Map 5: Census SAE of Poverty at the Locality Level ................................................................................... 35 Map 6: Census SAE of Number of Poor at the Locality Level...................................................................... 36 Map 7: Census SAE Gini at the Locality Level ............................................................................................. 37 v 1. Introduction Sudan is a country of considerable potential. Even after secession from South Sudan, it has abundant fertile land and livestock and a strategic market location. It has the potential to be a significant economic hub, lying as it does at the intersection of Sub-Saharan Africa and the Middle East. Moreover, Sudan is bordered by seven countries—the Arab Republic of Egypt, Eritrea, Ethiopia, South Sudan, the Central African Republic, Chad, and Libya—with four of them landlocked. The country’s strategic geographical location makes its political and economic success critical to the region; its failure would have significant detrimental implications for north, east, and central Africa. Sudan has suffered from political instability and conflict for five of the six decades since its independence in 1956. The current situation is precarious, given ongoing internal and external challenges. Sudan is home to about 40.53 million people (UNFPA 2017) with about two-thirds living in rural areas and about 60 percent of the population below the age of 25 years. During 1999–2011, Sudan had a decade of high real economic growth rates, driven by oil production and exportation. After the discovery of oil in 1999, the size of Sudan’s economy grew exponentially from US$12 billion in 1999 to US$65 billion in 2011—a 5.8 percent annual average growth. Over the same period, per capita income increased from US$934 to US$1,361 (constant 2010 U.S. dollar), raising Sudan to lower-middle-income status. Government revenue increased from 10 percent of gross domestic product (GDP) to 18 percent of GDP. The loss of oil production in 2011 brought about a deceleration of the economy but no recession. The economy continued to grow at a respectable 4.1 percent on average during 2012–17. Sudan posts very poor human development indicators for its level of GDP. In 2018, it ranked 139 out of 157 countries according to the World Bank Human Capital Index (HCI)1 and 167 out of 189 countries based on the Human Development Index (HDI).2 It did not meet the 2015 Millennium Development Goals (MDGs) and its progress lags on many fronts compared to its neighbors and to the Sub-Saharan African average. Education and health indicators remain low and vary markedly across states, gender, and poverty levels. The gross primary school enrolment rate is only 70 percent (below the target of universal coverage), with substantial disparities across states, urban/rural areas, and gender. The under-5 mortality rate of 68 deaths per 1,000 births in 2014 is still higher than the 2015 MDG target of 41 per 1,000 births. This means that a lot of effort will be needed to achieve the 2030 Sustainable Development Goal target of 25 deaths per 1,000 births. Similarly, infant mortality rate and maternal mortality remain far higher than the Sustainable Development Goal targets. 1 https://databank.worldbank.org/data/download/hci/HCI_2pager_SDN.pdf. 2 http://www.hdr.undp.org/en/composite/HDI. 1 Because of the country’s dire economic and financial situation, over one-third of Sudanese remain poor. In 2014/15, official estimates set the national poverty rate at 36.1 percent, indicating that some 13.4 million people were poor (CBS 2017). The official poverty rate is higher in urban areas (37.3 percent) than in rural areas (35.5 percent). There are marked spatial disparities in poverty incidence. Two-thirds of the population lives in rural areas. Disparities were also pronounced across states (Map 1). For instance, at about 67 percent, the incidence of poverty in either Central Darfur state or South Kordofan state was nearly five times higher than in the Northern state and double than in Khartoum state. The official Gini coefficient of 29.2 percent indicates that inequality was moderate compared to other Sub-Saharan African countries and in line with Middle East and North African countries. Map 1: Direct Estimates at the State Level Source: 2014/15 National Household Budget and Poverty Survey (NHBPS). 2 2. Methodology and Data 2.1. Methodology The small area estimation (SAE) methodology has gained widespread popularity among development practitioners around the world. This methodology assigns consumption levels to census households based on a consumption model estimated from the household survey. The consumption model includes explanatory variables—for example, household and individual characteristics—that are statistically identical in both the census and the household survey. The consumption expenditures of the census households are imputed by applying the estimated coefficients to the variables common to the survey and the census data. Poverty and inequality statistics for small areas are then calculated based on the imputed consumption of census households. Several poverty mapping methods have been used and documented by Bigman and Deichmann (2000). However, the selection of a specific poverty mapping methodology is a critical first step in deriving a poverty map. The SAE method developed by Elbers, Lanjouw, and Lanjouw (2003)— henceforth referred to as ELL––has acquired wide recognition among development practitioners around the world and is preferred within the World Bank when sufficient data are available (Mungai, Nguyen, and Pradhan 2018). ELL has been chosen to estimate parameters in all maps in this report. As input, ELL uses household-level data from the 2008 Population and Housing Census and the NHBPS of Sudan. As a unit-level model, it uses detailed income or consumption information at the household level combined with observable characteristics of the household to estimate welfare. As the household survey helps estimate parameters given a set of observables in the model, the census will serve to implement the simulation. Given that Sudan has done the census in 2008 and the NHBPS in 2014, the World Bank has access to these data, which makes this method appropriate for the exercise. The model procedures can be described in the following manner. Once the model parameters are estimated in the household survey data, they are applied to the census data to predict the welfare for households that possess the same characteristics. Then, poverty rates are calculated for each locality presented in the census. Errors may occur in the poverty rates calculation, but literature and experience help us conclude that the results are still accurate for informing policy choices (Bedi, Coudouel, and Simler 2007). The specificity of the ELL method is that the estimation of poverty incidence comes along with the estimation of the standard errors. This is not common for other poverty mapping methods. Notice that the standard errors estimate results from deriving the properties of the imputation errors obtained after using imputed consumption in the poverty estimates (Elbers, Lanjouw, and Lanjouw 2003). 3 The model formulation is as follows: ℎ = ′ℎ + ℎ (1), where ℎ is the log per capita consumption of household h residing in area c, ℎ refers to household and area/location characteristics, and ℎ = + ℎ , representing the residual, which is composed of the area component and the household component ℎ . and ℎ have 2) 2 2 expected values of zero and are independent of each other. It is assumed that ( = + . The estimation of variance parameters is done through Henderson’s method III, a commonly used estimator for the variance parameters of a nested error model (Henderson 1953; Searle, Casella, and McCulloch 1992). 2 ℎ A logistic transformation as a function of household and area characteristics [− 2 ] = ′ℎ + ℎ ℎ is used for the estimation of other variances such as the residual ℎ . However, heteroskedasticity is permitted so reestimation to get a general least squares (GLS) estimate of and of the variance-covariance matrix would be needed. As the main idea of SAE is the simulation, estimates are a means of that simulation. It can be written as: 1 ̂= ̃ ) (2), ∑ ℎ( =1 where ℎ() is a function that converts the vector y with (log) incomes for all households into a poverty measure (such as the head count rate), ̃ denotes the r-th simulated vector with the elements: ̃ + ̃ = ′ ̃ + ̃ℎ (3), and R is the number of simulations. This simulation approach is well fitted because measures of poverty and inequalities are nonlinear functions. According to Mungai, Nguyen, and Pradhan (2018), both the model parameters ̃ and the errors ̃ and ̃ℎ ̃ is drawn by are drawn from their estimated distributions for each simulation. reestimating the model parameters using the r-th bootstrap version of the survey sample. Otherwise, ̃ may be drawn from its estimated asymptotic distribution, referred to as parametric drawing. The parametric drawing is computationally fast but the true distribution of the estimator for the model parameter vector may differ from the asymptotic distribution. The use of bootstrapping, albeit more computationally intensive, provides a means of identifying the finite-sample distribution and is thus expected to provide more accurate results when the sample size is small. 4 The sample size in the NHBPS we use is large enough for the asymptotic results to apply, and for this reason we expect to see little to no difference between estimates obtained with parametric drawing and bootstrapping. Box 1: Step-by-step Summary of the Modelling Approach (1) Bootstrap the survey, unless parametric drawing of the model parameters is used. (2) Estimate by means of ordinary least squares (OLS) and extract the residuals. (3) Regress residuals from (2) on the area dummies (that is, estimate Fixed-effects [FE] model) and extract the residuals. 2 2 (4) Estimate the unconditional variance parameters of the nested error model ( and ) by applying the Henderson method III (Henderson 1953), which uses the residuals from both (2) and (3). (5) If heteroskedastic household errors are assumed, then (a) derive the estimates of the household errors by subtracting the area averages from the residuals (that is, deviations from the area mean residual), (b) apply a logistic transformation to the errors derived under (a) to obtain the left hand side (LFS) of the regression (also referred to as the alpha model) that will be used to predict the conditional variance of the household 2 2 component ℎ , denoted by ,ℎ , and (c) ensure that the unconditional variance is still equal to , that is, 2 2 [,ℎ ] = . 2 2 (6) Given the estimates of the unconditional variance and conditional variance ,ℎ , we may construct the covariance matrix Ω, which is used to obtain the GLS estimator for . (7) At this stage, we have the estimates for all the model parameters ̃ . Next, we draw the area errors and the household idiosyncratic errors (5) from their respective normal distributions with variances (8) We now have all we need to compute the round r simulated (log) household expenditure values for all households in the population census (9) With the simulated household income data, we can now compute the poverty and inequality measures as if the population census came with household income data from the start. (10) This yields a simulated poverty and inequality measure for each of the R simulation rounds. The average and standard deviation give us the poverty points estimate and the corresponding standard error, respectively. Source: Mungai, Nguyen, and Pradhan 2018. Another method that can be applied is the Empirical Best (EB) estimation. It assumes that for households sampled in area c, residuals ℎ = ℎ − ′ℎ are informative of the latent area error . Thus, if we are conditioning on the residuals observed for sampled households, it should enable us to tighten the distributions from which to simulate (Mungai, Nguyen, and Pradhan 2018). The EB can only be applied for the drawing of the area errors which have been sampled in the survey. Other areas will use the unconditional distribution called ELL-EB and match with the standard ELL. 5 2.2. Main Sources of Data 2.2.1 Census and NHBPS The poverty mapping exercise for Sudan combines data from the 2014/15 NHBPS and the 2008 household census. The 2008 Sudan Population and Housing Census was fielded during April 22 to May 6, 2008. The census provides comprehensive information on the household sociodemographic conditions, dwelling conditions, and individual characteristics of household members (for example, age, education, and marital and employment status), but it does not include information to construct consumption-based welfare measures. Sudan’s 2014/15 NHBPS consisted of three waves of data collection in November 2014, March 2015, and August 2015. The 2014/15 NHBPS collected consumption information used to calculate expenditure at the household level for all households. During the two 2015 waves, only consumption and expenditure data were collected, but the March 2015 round did not administer module five, which records nonfood consumption with a 12-month recall. The aim of this design was to explore and account for seasonality. The national poverty analysis exercise relied on a sample of 11,953 households, with consumption averaged over the three waves. However, inspection of the item-level consumption and expenditure records showed that 13,733 households were initially interviewed.3 The remaining 1,780 households, 13 percent of the initial sample, were dropped, mainly because they were not interviewed in either wave two or wave three (or both). Sampling weights were scaled up by the Central Bureau of Statistics (CBS), with different scaling factors applied across primary sampling units (PSUs). However, it is not clear exactly how this was done. Sudan’s CBS implemented the 2014/15 NHBPS with funding support by African Development Bank (AfDB). The lowest level of representativeness in the 2014/15 NHBPS was the state. The sampling frame for the 2014/15 NHBPS was the 2008 Population and Housing Census. 2.2.2 Matching NHBPS and Census Data The log per capita consumption forms the dependent variable of our models and is also used for the official measurement of poverty reported by the central statistical office. For constructing a unit-level model, the exercise relied on the NHBPS data on several household and personal characteristics—such as household composition, age, gender, and level of education—as well as dwelling characteristics, assets, and land ownership. The NHBPS data were combined with the data from the 2008 household census. 3 In addition to detailed expenditure and consumption data, the data sets obtained included household and enumeration area identifiers as well as information about the locality (state of residence and rural/urban locality). 6 The census data similarly cover several key household and individual characteristics, including (a) Demography: age/sex profiles, marital status, and household composition; (b) Educational attainment; (c) Information on dwellings: type of ownership, amenities, number and surface of rooms, type of sewerage facilities, and type of dwelling; and (d) Assets and land ownership; see Appendix A’ for the complete list of common variables. As the ELL setup is based on estimating a welfare model on the NHBPS data and applying it to the population census data for prediction, one of the important parts of the model setup is the congruence between the variables in the NHBPS and the census. As part of building a welfare model, like for The Gambia, a two-step process was undertaken: • Step 1. Compare the NHBPS and population census questionnaires to identify ‘candidate variables’ that exist both in the survey and the census and that are generated from identical or similar questions (see Appendix A’); and • Step 2. Compare the distributions of the ‘candidate variables’ identified in Step 1 to examine whether they appear to capture the same underlying phenomena or whether, despite similar questions, their empirical distributions differ in any important ways. Given that the goal of the model construction is to create a descriptive model which explains the variation in household consumption, the selection of candidate variables relies on a heuristic model of households’ consumption. Thus, the consumption pattern of the household is assumed to be a function of (a) The types of individuals in the household, for example, age of children, working-age adults, or elderly; and (b) Income-earning characteristics of the household, for example, highest level of education of the household members. In addition, while they are not determinants of income-earning capacity, the type of dwelling where the household resides or the types of assets the household possesses—for example, whether there is a bath or toilet in the dwelling—are also assumed to be able to describe or ‘reflect’ the income level of the household. Moreover, household income may also change across a given set of household characteristics or the location of the household, for example, rural versus urban, proximity to big cities, area with low or high employment rates, and so on. The above list is not unique or exhaustive, but the overlap between the survey and census questionnaires is the main constraint in the choice of the characteristics. 7 2.3. Modeling for Monetary Poverty As described in the previous paragraph, the following variables were chosen since they are common to the survey and the census: • Demographic characteristics. Gender; age; marital status; relationship to household head; household size; number of children, adults, and elderly in the household; and dependency ratio • Education. Education level of the household head, literacy, and highest level of education of any household member • Occupation. Employment status, occupation, and sector of employment of the household head. • Housing characteristics. Type of housing unit, land and dwelling information, ownership and occupancy status of dwelling, type of energy used, source of drinking water and electricity, and type of toilet • Productive and durable assets. Ownership of radio, television, personal computer, fan, air conditioner, refrigerator, motor vehicle, motorcycle, bicycle, canoe/boat, livestock, and poultry. Depending on the data, single or multiple regression models can be fitted. The single regression model assumes that there is only one model that describes the poverty phenomenon in the whole country. The link between the income or consumption for all households and their characteristics are uniform for every single household, no matter the region they are in. All parameters are equal. This type of model is not realistic and is highly biased for a country like Sudan where spatial heterogeneity is mainly dominant. This can be in terms of climate, geography, security, returns in education, the capacity of each region to manage its natural resources, the availability of a formal job market, industry, and main economic activity. The multiple regression models are a more convenient way to surpass the single model. Estimating a model for each region can be time consuming, but it offers good quality. First, the relationship between expenditure and the explanatory variables can differ throughout the country that induces more flexibility in the type of places where the unit is evolving. Second, it lessens the standard error of poverty prediction due to the error in modeling. Otherwise, introducing regional dummy variables in the regression can give similar results while having only one model. 8 Before an in-depth look at the modeling, let us describe the data. Table 1 displays the geographical distribution of the census4 and survey. Table 1: Geographic Distribution between the Census (2008) and the 2014/15 NHBPS Censusa Survey Sample Size Weighted Sample Size Weighted Number of households 917,453 5,316,971 11,953 6,001,018 Number of individuals 5,049,590 30,248,885 69,828 34,574,848 Male 2,629,262 15,289,254 35,081 17,401,614 Female 2,420,328 14,959,631 34,747 17,173,234 Regions 6 — 6 — States 15 — 18 — Counties 131 — 134 — Note: a. CBS provided a sample census data accounting for 16.6 percent of the total census population. So for any analysis one must weight the data to get the correct population size. In this mapping, we chose to start from the national-level regression and go down to the region- level regression models to better determine the forms of regression to adopt. However, region in the census and region in the survey are not directly comparable due to changes in the government boundaries. Hence, one region may differ in states as well as one state may differ in counties components. Therefore, getting the regional-level regression model has implied aggregation of the data from the locality level. While multiple regression can offer flexibility to the parameter across the region, it also causes a loss in degrees of freedom and there is a risk of overfitting.5 A solution that researchers recommended to avoid overfitting is that the sample size should be no smaller than 300 for each regression (Ahmed et al. 2014). To do this task, the means of candidate variables were manually compared between the two data sets. ‘Acceptable’ variables are included in the model selection, and the ‘non-acceptable’ variables are excluded. Criteria to define acceptable versus non-acceptable variables are based on the differences of means. Table 2 to Table 7 list the means of the variables evaluated at the household level by region and the significance of the test of mean equality (pvalue). Table 2: Comparison of Household Characteristics between the Census (2008) and the 2014/15 NHBPS, Weighted Average, Variables Used in Model for Region 1 Variables Description mean1 mean2 p-value Average dependency ratio (less than 18 + 65+ over adult depratio pop) 1.164 1.132 0.376 dwelling1 Household having dwelling: Tent 0.098 0.037 0.000 head_age25t64 Household having head age of 25 to 64 0.774 0.782 0.573 4 Sudan Census was done before the break-up. 5 The models are forced to explain and justify the noise in the data in a small sample. 9 Variables Description mean1 mean2 p-value head_edlevel3 Household whose head has education: Secondary 0.100 0.068 0.001 head_edlevel4 Household whose head has education: Tertiary 0.040 0.268 0.000 head_employer Household whose head is employer 0.062 0.035 0.000 head_male Household having male head 0.827 0.901 0.000 head_selfempl Household whose head is self-employed 0.352 0.351 0.919 hhsize_2 Household with 2 members 0.113 0.078 0.001 hhsize_3 Household with 3 members 0.126 0.115 0.306 hhsize_4 Household with 4 members 0.134 0.168 0.002 hhsize_5 Household with 5 members 0.133 0.166 0.003 hhsize_6 Household with 6 members 0.117 0.145 0.006 nrooms Average number of rooms 2.146 2.172 0.544 sector1_share Share of member of household in agriculture 0.191 0.139 0.000 toilet1 Household having house toilet: Pit latrine private 0.505 0.864 0.000 toilet4 Household having house toilet: Flush toilet shared 0.005 0.072 0.000 urban Share of urban population 0.215 0.226 0.374 Source: Authors’ calculations based on the 2014/15 NHBPS and the 2008 Population and Housing Census. Table 3: Comparison of Household Characteristics between the Census (2008) and the 2014/15 NHBPS, Weighted Average, Variables Used in Model for Region 2 Variables Description mean1 mean2 p-value charcoal Household cooking with fuel: Charcoal 0.131 0.260 0.000 child_2 Household with two children 0.601 0.576 0.033 Average dependency ratio (less than 18 + 65+ over depratio adult pop) 1.264 1.256 0.764 dwelling3 Household having dwelling: Tukul/gottiya of mud 0.108 0.306 0.000 Household having access to electricity as source of elec lighting 0.341 0.462 0.000 Household having access to electricity as source of elec_m_county lighting (county/locality mean) 0.357 0.406 0.000 gas Household cooking with fuel: Gas 0.081 0.312 0.000 head_edlevel1 Household whose head has education: None 0.024 0.276 0.000 head_edlevel2 Household whose head has education: Primary 0.181 0.090 0.000 head_edlevel3 Household whose head has education: Secondary 0.029 0.044 0.000 head_employed Household whose head is employed 0.781 0.895 0.000 head_employer Household whose head is employer 0.091 0.087 0.530 head_unpaid Household whose head is unpaid 0.076 0.004 0.000 hhsize_2 Household with 2 members 0.086 0.107 0.002 hhsize_3 Household with 3 members 0.116 0.136 0.008 hhsize_4 Household with 4 members 0.142 0.144 0.785 hhsize_5 Household with 5 members 0.154 0.150 0.682 hhsize_6 Household with 6 members 0.137 0.150 0.115 10 Variables Description mean1 mean2 p-value nrooms Average number of rooms 1.517 1.637 0.000 Share of member having completed primary pri_abv_share education and above 0.048 0.625 0.000 sector1_share Share of member of household in agriculture 0.277 0.179 0.000 sector3_share Share of member of household in services 0.143 0.291 0.000 sum_age1t14_m_count Average number of children of ages 1 to 15 years y (county/locality mean) 2.260 2.198 0.000 sum_edlevel1_m_count Average number of persons in a household with y education level: None (county/locality mean) 0.137 1.643 0.000 sum_selfempl_m_count Average number of persons in a household self- y employed (county/locality mean) 0.611 0.551 0.000 Average number of persons in a household unpaid sum_unpaid_m_county (county/locality mean) 0.210 0.165 0.000 toilet2 Household having house toilet: Pit latrine shared 0.026 0.081 0.000 urban Share of urban population 0.172 0.373 0.000 Source: Authors’ calculations based on the 2014/15 NHBPS and the 2008 Population and Housing Census. Table 4: Comparison of Household Characteristics between the Census (2008) and the 2014/15 NHBPS, Weighted Average, Variables Used in Model for Region 3 Variables Description mean1 mean2 p-value child_2 Household with two children 0.537 0.547 0.516 Average dependency ratio (less than 18 + 65+ over depratio adult pop) 1.003 1.096 0.007 dwelling4 Household having dwelling: Tukul/gottiya of sticks 0.009 0.030 0.000 dwelling5 Household having dwelling: Flat or apartment 0.014 0.020 0.122 Share of individuals of age 18 to 64 years having everattend_share ever attended school 0.987 0.989 0.504 head_edlevel4 Household whose head has education: Tertiary 0.157 0.200 0.000 head_male Household having male head 0.787 0.846 0.000 Household having marital status of the head: head_martial3 Widowed 0.050 0.075 0.000 head_selfempl Household whose head is self-employed 0.223 0.145 0.000 hhsize_2 Household with 2 members 0.086 0.080 0.486 hhsize_3 Household with 3 members 0.098 0.113 0.133 hhsize_4 Household with 4 members 0.115 0.139 0.024 hhsize_5 Household with 5 members 0.122 0.175 0.000 hhsize_6 Household with 6 members 0.117 0.172 0.000 literacy_share Share of member literate in a household 0.776 0.876 0.000 Share of member having completed primary pri_abv_share education and above 0.386 0.590 0.000 Share of member having completed secondary sec_abv_share education and above 0.158 0.364 0.000 sum_edlevel3_m_count Average number of persons in a household with y education level: Secondary (county/locality mean) 0.887 0.601 0.000 tenure2 Household having house tenure: Rented 0.250 0.230 0.162 toilet1 Household having house toilet: Pit latrine private 0.601 0.772 0.000 11 Variables Description mean1 mean2 p-value urban Share of urban population 0.807 0.795 0.345 Source: Authors’ calculations based on the 2014/15 NHBPS and the 2008 Population and Housing Census. Table 5: Comparison of Household Characteristics between the Census (2008) and the 2014/15 NHBPS, Weighted Average, Variables Used in Model for Region 4 Variables Description mean1 mean2 p-value Household cooking with fuel: Charcoal charcoal_m_county (county/locality mean) 0.151 0.194 0.000 Average dependency ratio (less than 18 + 65+ over depratio adult pop) 1.407 1.414 0.763 dwelling1 Household having dwelling: Tent 0.089 0.241 0.000 Household having dwelling: Tukul/gottiya of mud dwelling3_m_county (county/locality mean) 0.121 0.647 0.000 head_edlevel4 Household whose head has education: Tertiary 0.034 0.367 0.000 head_employee Household whose head is employee 0.256 0.439 0.000 head_employer Household whose head is employer 0.059 0.070 0.015 Household having marital status of the head: head_martial2 Married 0.869 0.905 0.000 hhsize_2 Household with 2 members 0.086 0.067 0.000 hhsize_3 Household with 3 members 0.109 0.110 0.860 hhsize_4 Household with 4 members 0.128 0.146 0.005 hhsize_5 Household with 5 members 0.132 0.158 0.000 hhsize_6 Household with 6 members 0.125 0.142 0.007 literacy_share Share of member literate in a household 0.507 0.630 0.000 nrooms Average number of rooms 1.833 1.988 0.000 Share of member having completed primary pri_abv_share education and above 0.141 0.539 0.000 sector1_share Share of member of household in agriculture 0.164 0.187 0.000 toilet1 Household having house toilet: Pit latrine private 0.324 0.689 0.000 toilet2 Household having house toilet: Pit latrine shared 0.105 0.045 0.000 urban Share of urban population 0.195 0.285 0.000 Household having access to drinking water (state water_m_state mean) 0.485 0.370 0.000 Source: Authors’ calculations based on the 2014/15 NHBPS and the 2008 Population and Housing Census. Table 6: Comparison of Household Characteristics between the Census (2008) and the 2014/15 NHBPS, Weighted Average, Variables Used in Model for Region 5 Variables Description mean1 mean2 p-value bicycle Household with bicycle 0.062 0.103 0.000 charcoal Household cooking with fuel: Charcoal 0.065 0.165 0.000 child_1 Household with one child 0.807 0.811 0.644 child_3p Household with three and more children 0.459 0.519 0.000 Average dependency ratio (less than 18 + 65+ over depratio adult pop) 1.485 1.558 0.008 12 Variables Description mean1 mean2 p-value dwelling3 Household having dwelling: Tukul/gottiya of mud 0.105 0.299 0.000 Share of individuals of age 18 to 64 years having everattend_share ever attended school 0.972 0.983 0.002 fan Household with fan 0.014 0.062 0.000 gas Household cooking with fuel: Gas 0.022 0.069 0.000 head_edlevel1 Household whose head has education: None 0.048 0.274 0.000 head_edlevel3 Household whose head has education: Secondary 0.022 0.034 0.000 head_male Household having male head 0.778 0.873 0.000 Household having marital status of the head: head_martial3 Widowed 0.050 0.058 0.099 head_selfempl Household whose head is self-employed 0.428 0.533 0.000 hhsize_2 Household with 2 members 0.096 0.061 0.000 hhsize_3 Household with 3 members 0.130 0.099 0.000 hhsize_4 Household with 4 members 0.150 0.120 0.000 hhsize_5 Household with 5 members 0.152 0.144 0.350 hhsize_6 Household with 6 members 0.140 0.128 0.106 motor Household with motor 0.016 0.028 0.000 phone Household with phone 0.139 0.767 0.000 sector3_share Share of member of household in services 0.144 0.176 0.000 Average number of persons in a household in sum_sector1_m_state sector: Agriculture (state mean) 0.826 0.841 0.000 toilet4 Household having house toilet: Flush toilet shared 0.001 0.341 0.000 tv Household with TV 0.051 0.187 0.000 urban Share of urban population 0.083 0.222 0.000 water Household having access to drinking water 0.231 0.148 0.000 Source: Authors’ calculations based on the 2014/15 NHBPS and the 2008 Population and Housing Census. Table 7: Comparison of Household Characteristics between the Census (2008) and the 2014/15 NHBPS, Weighted Average, Variables Used in Model for Region 6 Variables Description mean1 mean2 p-value child_1 Household with one child 0.832 0.869 0.000 child_2 Household with two children 0.664 0.727 0.000 computer Household with computer 0.001 0.010 0.000 Average dependency ratio (less than 18 + 65+ over depratio adult pop) 1.508 1.827 0.000 Household having access to electricity as source of elec lighting 0.065 0.145 0.000 Share of individuals of age 18 to 64 years having everattend_share ever attended school 0.974 0.971 0.200 firewood Household cooking with fuel: Firewood 0.974 0.858 0.000 gas Household cooking with fuel: Gas 0.002 0.008 0.000 head_literacy Household whose head is able to read and write 0.229 0.561 0.000 13 Variables Description mean1 mean2 p-value Household having marital status of the head: head_martial2 Married 0.878 0.846 0.000 hhsize_2 Household with 2 members 0.084 0.058 0.000 hhsize_3 Household with 3 members 0.128 0.101 0.000 hhsize_4 Household with 4 members 0.157 0.132 0.000 hhsize_5 Household with 5 members 0.159 0.141 0.004 hhsize_6 Household with 6 members 0.133 0.145 0.042 motor Household with motor 0.007 0.017 0.000 motorcycle Household with motorcycle 0.005 0.046 0.000 phone Household with phone 0.070 0.638 0.000 Share of member having completed primary pri_abv_share education and above 0.029 0.604 0.000 radio Household with radio 0.482 0.256 0.000 refri Household with refrigerator 0.005 0.031 0.000 sector3_share Share of member of household in services 0.151 0.220 0.000 tenure2 Household having house tenure: Rented 0.014 0.071 0.000 toilet2 Household having house toilet: Pit latrine shared 0.032 0.044 0.000 tv Household with TV 0.021 0.127 0.000 urban Share of urban population 0.044 0.214 0.000 Household having access to drinking water water_m_county (county/locality mean) 0.166 0.181 0.000 Source: Authors’ calculations based on the 2014/15 NHBPS and the 2008 Population and Housing Census. 2.4. Technical Challenges Evolving Administrative Boundaries and Classifications Evolving administrative boundaries posed a technical challenge during this poverty mapping exercise. After the country split on July 2011, Sudan reviewed its administrative unit. The divisions, regions, states, and counties did not change but their composition did. For example, new states have been created in the Kordofan and Darfur regions. New states have been created such as in the Kordofan and Darfur regions. Kordofan, which earlier had two states, now has three. Darfur had three states, and now it has five. Since the census was done before the split and the NHBPS after, it led to a change in the definition of the mapping. To tackle this issue, census localities were redefined according to the subdivision used in the NHBPS. It must be noted however that some of the localities in the census may not have any match in the survey, whether they are merged with other localities or replaced with a new one. 14 Only the localities having a match in the survey and those coming from the survey are used for the simulation; localities that were used in only the census are not needed. So even though completed in majority, simulated poverty data in this mapping are not exhaustive for Sudan. Other technical issues that we needed to overcome were the differences between the census and survey classifications. This was especially true of the education and employment sections. The education level section, ‘currently attending’ or ‘was attending’, has 17 categories but the content differs in the two questionnaires. It leads to confusion in determining the exact highest level to be taken in account. So, for more aggregated classification, only the four categories— None, Primary, Secondary, and Tertiary—are used. With regard to the employment section, the reference period for the variable asking ‘Work’ is different between the survey and the census; 10 days for the survey and 7 days for the census. The screening question was different as well. This may involve differences in comments that we need to be careful about. However, removing this variable would make the model weak as employment is a determinant for income and, consequently, for poverty. 15 3. Constructing the 2014/15 Sudan Poverty Maps 3.1. Model Selection In addition to the manual selection of the variables to exclude at the first stage in the regression, the model selection borrowed the automated procedures performed in the poverty mapping for The Gambia (Mungai, Nguyen, and Pradhan 2018). The main advantages of this method are that it minimizes overfitting by incorporating the degrees of freedom into the evaluation. The modified stepwise procedure involves using the variance inflation factor (VIF), sequential removal of one variable at a time from GLS estimates, and rerunning it stepwise. The process is repeated until all variables in GLS estimate are significant. Technically, it first removes variables for which p-value is greater than 0.2 one by one, and then the variable with VIF more than 5. This last process prevents the multicollinearity between variables. Thompson (1995) can offer more detail on this procedure for modeling. Stepwise Akaike Information Criteria (AIC) is then undertaken after the default model to limit overfitting. AIC is the information-based criteria that can be performed using ‘vselect’ in Stata. The score estimates the expected relative distance between the fitted model and the unknown true relationship. The purpose of this procedure is to minimize the AIC. Naming r, the number of parameters in the model, the score for AIC is formulated as = −2 log(ℎ) + 2. See Lindsey and Sheather (2010) for further details on information-based selection and the vselect package. After using the implementation of the ELL methods in Stata 15 to build the model and following the validation process we just described, final models are specific for each region. The initial welfare models corresponding to equation (1) are presented in Tables 8 to 13, for each region. The adjusted R-squared for regional models is moderately high, ranging from 0.46 to 0.65. This means that the independent variables in the chosen model explain the variation on welfare moderately well. Variable means at the region and district levels obtained from the census are introduced to the model to improve precision. It may moderate the unexplained variation in income due to location. With the inclusion of these variables, the ratios of the variance of over regional models’ Mean Square Error (MSE) are from 4 percent to 15 percent. The low ratio shows the key role the variables play in improving the precision of the estimates. The estimated coefficients in the previous section serve as inputs to estimate the first part of the ̂ ) by combining coefficients with the census variables. However, vectors of equation (′ℎ 16 disturbances for households are still unknown and so must be estimated. Thus, the error decomposition is done through Henderson’s method III. The coefficients are obtained by bootstrapped samples of the NHBPS data. The final model chosen is where and are drawn from a normal distribution with their respective variance structures. Finally, EB methods are chosen since these incorporate more information and are expected to provide a better fit. The model selection used was the stepwise with VIF using AIC criteria based on the comparison of poverty estimates from the survey and the census at the national and regional levels. Tables 8 to 13 show the final regional model estimate (bGLS) that is preferred compared to the OLS model. Each also provides results of the disturbances selection process. All variables in final GLS estimate are significant. Table 8: Model Estimates Based on the 2014/15 NHBPS - ‘Northern’ (Beta Model, Region 1) bOLS bGLS Standard Coefficient Coefficient Standard Error Error Depratio −0.055*** 0.011 −0.040*** 0.010 dwelling1 −0.264*** 0.065 −0.243*** 0.077 head_age25t64 −0.113*** 0.030 −0.107*** 0.028 head_edlevel3 0.183*** 0.045 0.203*** 0.041 head_edlevel4 −0.121*** 0.027 −0.100*** 0.025 head_employer 0.167*** 0.059 0.177*** 0.052 head_male −0.180*** 0.038 −0.156*** 0.036 head_selfempl 0.111*** 0.024 0.129*** 0.022 hhsize_2 0.844*** 0.050 0.881*** 0.048 hhsize_3 0.591*** 0.042 0.655*** 0.042 hhsize_4 0.419*** 0.035 0.459*** 0.030 hhsize_5 0.246*** 0.034 0.289*** 0.027 hhsize_6 0.161*** 0.034 0.192*** 0.029 Nrooms 0.054*** 0.013 0.074*** 0.011 sector1_share −0.118** 0.053 −0.134*** 0.043 toilet1 −0.192*** 0.049 −0.149*** 0.044 toilet4 −0.191*** 0.068 −0.212*** 0.062 Urban −0.257*** 0.029 −0.244*** 0.025 _cons 9.220*** 0.082 9.057*** 0.078 Number of observations 1,002 Error decomposition ELL Adjusted R-squared 0.497 EB methods No Sigma ETA sq. 0.004 Beta drawing Bootstrapped Ratio of sigma eta sq over MSE 0.039 Eta drawing method Normal Variance of epsilon 0.099 Epsilon drawing method Normal Sampling variance of Sigma eta sq. 6E-06 Alpha model Yes Source: Authors’ calculations based on the 2014/15 NHBPS. Note: Model settings - Alpha model; Error decomposition - ELL; Bootstrapped with normal distribution; EB methods. ***p < 0.01; **p < 0.05; *p < 0.1. 17 Table 9: Model Estimates Based on the 2014/15 NHBPS - ‘Eastern’ (Beta Model, Region 2) bOLS bGLS Standard Coefficient Coefficient Standard Error Error Charcoal 0.073*** 0.025 0.047** 0.022 child_2 −0.056** 0.027 −0.061** 0.024 Depratio −0.026** 0.010 −0.032*** 0.008 dwelling3 0.099*** 0.023 0.109*** 0.020 Elec 0.113*** 0.026 0.106*** 0.022 elec_m_county 0.259*** 0.061 0.318*** 0.118 Gas 0.167*** 0.031 0.139*** 0.026 head_edlevel1 0.070*** 0.024 0.050** 0.020 head_edlevel2 0.098*** 0.033 0.057** 0.025 head_edlevel3 0.155*** 0.046 0.129*** 0.034 head_employed −0.076** 0.030 −0.066** 0.029 head_employer 0.093*** 0.031 0.087*** 0.028 head_unpaid 0.224* 0.127 0.178** 0.075 hhsize_2 0.882*** 0.039 0.901*** 0.032 hhsize_3 0.597*** 0.036 0.609*** 0.033 hhsize_4 0.442*** 0.030 0.448*** 0.025 hhsize_5 0.336*** 0.028 0.345*** 0.023 hhsize_6 0.200*** 0.027 0.213*** 0.024 Nrooms 0.057*** 0.012 0.054*** 0.009 pri_abv_share −0.070*** 0.027 −0.052** 0.023 sector1_share 0.094** 0.041 0.087** 0.035 sector3_share 0.147*** 0.039 0.169*** 0.034 sum_age1t14_m_county 0.361*** 0.032 0.364*** 0.064 sum_edlevel1_m_county −0.098*** 0.027 −0.117** 0.048 sum_selfempl_m_county −0.443*** 0.059 −0.345*** 0.119 sum_unpaid_m_county −0.336*** 0.089 −0.435** 0.170 toilet2 0.153*** 0.034 0.146*** 0.033 Urban −0.348*** 0.027 −0.347*** 0.027 _cons 8.027*** 0.087 8.028*** 0.162 Number of observations 1,633 Error decomposition ELL Adjusted R-squared 0.582 EB methods No Sigma ETA sq. 0.008 Beta drawing Bootstrapped Ratio of sigma eta sq over MSE 0.077 Eta drawing method Normal Variance of epsilon 0.094 Epsilon drawing method Normal Sampling variance of Sigma eta sq. 8.899e-06 Alpha model Yes Source: Authors’ calculations based on the 2014/15 NHBPS. Note: Model settings - Alpha model; Error decomposition - ELL; Bootstrapped with normal distribution; EB methods. ***p < 0.01; **p < 0.05; *p < 0.1. Table 10: Model Estimates Based on the 2014/15 NHBPS - ‘Khartoum’ (Beta Model, Region 3) bOLS bGLS Standard Coefficient Coefficient Standard Error Error child_2 −0.127*** 0.035 −0.105*** 0.032 Depratio −0.058*** 0.017 −0.059*** 0.016 dwelling4 0.402*** 0.070 0.432*** 0.052 18 bOLS bGLS Standard Coefficient Coefficient Standard Error Error dwelling5 0.296*** 0.084 0.324*** 0.075 everattend_share −0.776*** 0.125 −0.784*** 0.233 head_edlevel4 −0.141*** 0.036 −0.105*** 0.035 head_male −0.078** 0.040 −0.075** 0.037 head_martial3 −0.125** 0.054 −0.124** 0.050 head_selfempl 0.100*** 0.033 0.075** 0.030 hhsize_2 0.837*** 0.050 0.868*** 0.045 hhsize_3 0.622*** 0.044 0.648*** 0.040 hhsize_4 0.455*** 0.038 0.472*** 0.035 hhsize_5 0.324*** 0.034 0.336*** 0.032 hhsize_6 0.234*** 0.034 0.238*** 0.032 literacy_share 0.239*** 0.058 0.255*** 0.060 pri_abv_share 0.186*** 0.043 0.151*** 0.036 sec_abv_share 0.195*** 0.048 0.228*** 0.041 sum_edlevel3_m_county 0.354*** 0.050 0.343** 0.144 tenure2 −0.055** 0.028 −0.078*** 0.026 toilet1 −0.255*** 0.033 −0.243*** 0.029 Urban −0.249*** 0.029 −0.187*** 0.029 _cons 9.460*** 0.131 9.383*** 0.253 Number of observations 930 Error decomposition H3 Adjusted R-squared 0.652 EB methods Yes Sigma ETA sq. 0.009 Beta drawing Bootstrapped Ratio of sigma eta sq over MSE 0.079 Eta drawing method Normal Variance of epsilon 0.111 Epsilon drawing method Normal Alpha model Yes Source: Authors’ calculations based on the 2014/15 NHBPS. Note: Model settings - Alpha model; Error decomposition - ELL; Bootstrapped with normal distribution; EB methods. ***p < 0.01; **p < 0.05; *p < 0.1. Table 11: Model Estimates Based on the 2014/15 NHBPS - ‘Central’ (Beta Model, Region 4) bOLS bGLS Standard Coefficient Coefficient Standard Error Error charcoal_m_county 0.224*** 0.061 0.282** 0.129 Depratio −0.043** 0.007 −0.034** 0.007 dwelling1 −0.141** 0.022 −0.133*** 0.023 dwelling3_m_county −0.305*** 0.038 −0.237*** 0.071 head_edlevel4 −0.072*** 0.022 −0.076*** 0.020 head_employee −0.060*** 0.016 −0.066*** 0.015 head_employer 0.128*** 0.030 0.113*** 0.029 head_martial2 0.085*** 0.026 0.105** 0.024 hhsize_2 1.041*** 0.035 1.078** 0.034 hhsize_3 0.689*** 0.029 0.710*** 0.028 hhsize_4 0.457** 0.025 0.472** 0.022 hhsize_5 0.288*** 0.024 0.304*** 0.023 hhsize_6 0.200* 0.024 0.216** 0.023 literacy_share 0.110*** 0.028 0.098*** 0.027 19 bOLS bGLS Standard Coefficient Coefficient Standard Error Error Nrooms 0.054*** 0.009 0.062*** 0.008 pri_abv_share 0.150*** 0.024 0.141*** 0.022 sector1_share −0.085*** 0.031 −0.072*** 0.029 toilet1 0.072*** 0.019 0.066*** 0.020 toilet2 0.223*** 0.039 0.236*** 0.042 Urban −0.228*** 0.019 −0.229** 0.019 water_m_state 0.508** 0.062 0.486** 0.125 _cons 8.384*** 0.057 8.287*** 0.074 Number of observations 2,708 Error decomposition H3 Adjusted R-squared 0.566 EB methods Yes Sigma ETA sq. 0.004 Beta drawing Bootstrapped Ratio of sigma eta sq over MSE 0.039 Eta drawing method Normal Variance of epsilon 0.096 Epsilon drawing method Normal Alpha model Yes Source: Authors’ calculations based on the 2014/15 NHBPS. Note: Model settings - Alpha model; Error decomposition - H3; Bootstrapped with normal distribution; EB methods. ***p < 0.01; **p < 0.05; *p < 0.1. Table 12: Model Estimates Based on the 2014/15 NHBPS - ‘Kordofan’ (Beta Model, Region 5) bOLS bGLS Standard Standard Coefficient Coefficient Error Error Bicycle −0.140*** 0.029 −0.097*** 0.030 Charcoal 0.174*** 0.029 0.124*** 0.028 child_1 −0.138*** 0.029 −0.131*** 0.027 child_3p −0.108*** 0.028 −0.121*** 0.027 Depratio −0.034*** 0.009 −0.033*** 0.009 dwelling3 0.068** 0.027 0.097*** 0.027 everattend_share −0.605*** 0.075 −0.594*** 0.071 Fan 0.077 0.047 0.128*** 0.046 Gas 0.197*** 0.042 0.139*** 0.041 head_edlevel1 0.079*** 0.021 0.068*** 0.020 head_edlevel3 0.080 0.053 0.086* 0.050 head_male −0.137*** 0.032 −0.104*** 0.030 head_martial3 −0.095** 0.047 −0.092** 0.044 head_selfempl 0.079*** 0.019 0.041** 0.018 hhsize_2 0.685*** 0.044 0.693*** 0.042 hhsize_3 0.570*** 0.036 0.562*** 0.034 hhsize_4 0.307*** 0.034 0.307*** 0.032 hhsize_5 0.248*** 0.028 0.246*** 0.026 hhsize_6 0.120*** 0.028 0.132*** 0.027 Motor 0.314*** 0.053 0.306*** 0.050 Phone 0.061*** 0.022 0.047** 0.021 sector3_share 0.106*** 0.037 0.104*** 0.035 sum_sector1_m_state 0.076* 0.044 0.256** 0.129 toilet4 −0.036* 0.021 −0.043** 0.020 Tv 0.099*** 0.032 0.090*** 0.031 20 bOLS bGLS Standard Standard Coefficient Coefficient Error Error Urban −0.309*** 0.032 −0.259*** 0.033 Water 0.094*** 0.027 0.041 0.026 _cons 9.285*** 0.092 9.124*** 0.148 Number of observations 2,149 Error decomposition ELL Adjusted R-squared 0.511 EB methods No Sigma ETA sq. 0.014 Beta drawing Parametric Ratio of sigma eta sq over MSE 0.112 Eta drawing method Normal Variance of epsilon 0.114 Epsilon drawing method Normal Sampling variance of Sigma eta sq. 0.00008 Alpha model No Source: Authors’ calculations based on the 2014/15 NHBPS. Note: Model settings - Alpha model; Error decomposition - ELL; Parametric with normal distribution; EB methods. *** ***p < 0.01; **p < 0.05; *p < 0.1. Table 13: Model Estimates Based on the 2014/15 NHBPS – ‘Darfur’ (Beta Model, Region 6) bOLS bGLS Standard Coefficient Coefficient Standard Error Error child_1 −0.114*** 0.035 −0.108*** 0.032 child_2 −0.108*** 0.031 −0.105*** 0.029 Computer 0.231*** 0.087 0.252*** 0.081 Depratio −0.017** 0.007 −0.022*** 0.007 Elec 0.107*** 0.039 0.146*** 0.036 everattend_share −0.307*** 0.058 −0.281*** 0.054 Firewood −0.112*** 0.032 −0.128*** 0.030 Gas 0.254*** 0.090 0.245*** 0.084 head_literacy 0.080*** 0.022 0.085*** 0.020 head_martial2 −0.076*** 0.026 −0.076*** 0.024 hhsize_2 0.612*** 0.045 0.641*** 0.042 hhsize_3 0.492*** 0.036 0.508*** 0.033 hhsize_4 0.385*** 0.028 0.403*** 0.027 hhsize_5 0.243*** 0.027 0.247*** 0.025 hhsize_6 0.182*** 0.026 0.177*** 0.024 Motor 0.283*** 0.066 0.230*** 0.061 Motorcycle 0.110*** 0.042 0.107*** 0.039 Phone 0.119*** 0.020 0.120*** 0.019 pri_abv_share 0.051* 0.027 0.046* 0.025 Radio 0.135*** 0.020 0.092*** 0.019 Refri 0.075 0.056 0.093* 0.052 sector3_share 0.101*** 0.030 0.081*** 0.029 tenure2 0.087*** 0.034 0.113*** 0.031 toilet2 0.140*** 0.046 0.133*** 0.043 Tv 0.096** 0.040 0.078** 0.037 Urban −0.228*** 0.026 −0.245*** 0.028 water_m_county 0.597*** 0.043 0.555*** 0.156 _cons 8.770*** 0.065 8.800*** 0.074 Number of observations 3,444 Error decomposition H3 Adjusted R-squared 0.465 EB methods Yes 21 bOLS bGLS Standard Coefficient Coefficient Standard Error Error Sigma ETA sq. 0.023 Beta drawing Bootstrapped Ratio of sigma eta sq over MSE 0.155 Eta drawing method Normal Variance of epsilon 0.129 Epsilon drawing method Normal Alpha model No Source: Authors’ calculations based on the 2014/15 NHBPS. Note: Model settings - Alpha model; Error decomposition - H3; Bootstrapped with normal distribution; EB methods. ***p < 0.01; **p < 0.05; *p < 0.1. A visual assessment was conducted to compare the predicted and simulated consumption distributions as displayed in Figures 1 and 2. The results are based on various training samples— beginning with 10 percent of the sample and continuing up to 90 percent—at each region to ensure the robustness of the approach and the resulting distributions. This demonstrates a high level of statistical precision. However, this precision level declines as the degree of spatial disaggregation increases. This approach should be supplemented with complementary sources of information if further lower-level disaggregation is envisaged, but this should be done with a lot of caution. Figure 1: Distributions with Actual and Imputed Testing Sample 1 kdensity lnpcexp kdensity yhat10 kdensity yhat20 .8 kdensity yhat30 kdensity yhat40 kdensity yhat50 .6 kdensity yhat60 Density kdensity yhat70 kdensity yhat80 .4 kdensity yhat90 .2 0 6 8 10 12 log of welfare Source: Authors' calculation Source: Authors’ calculations based on the 2014/15 NHBPS. 22 Figure 2: Weighted Ratio Mean Square Error of Out sample for Sudan and its Regions .4 National Northern Eastern Khartoum .3 Central Kordofan Darfur .2 .1 0 10 20 30 40 50 60 70 80 90 x% training data insample Source: Authors' calculation Source: Authors’ calculations based on the 2014/15 NHBPS. 3.2. Level of Disaggregation The clustering used for estimations is at the PSU level and the poverty mapping results are based on survey direct estimates. To measure the share of the poor, the poverty line of SDG 5,109.78 per year per capita is used. Table 14 displays the poverty head count for direct and poverty mapping at the national level and across the region. One may notice that the World Bank’s poverty rate6 is different from these numbers. The reason is that the World Bank poverty rate of 46.5 per cent was estimated on the basis of Survey 2009 with the poverty line at SDG 114 per month per capita. However, the poverty rate in this report considers the national poverty rate estimated by the CBS through the 2014/15 NHBPS. The results of the mapping are similar to the estimates obtained from the NHBPS. At the national level, direct and small area estimates provide a good match because they differ at about +1 percentage point. Furthermore, the differences between the survey and small area estimates across regions are significantly low. 6 https://data.worldbank.org/indicator/si.pov.nahc 23 Table 14: Poverty Estimates from Survey (Observed) and the Census (SAE) Survey Census 95% Head Standard No. of No. of Head Standard Confidence Count Error Households Individuals Count Error Interval Sudan 36.14 0.44 35.28 37.00 917,453 29,757,647 37.47 0.81 Northern 17.14 1.19 14.80 19.47 38,895 1,730,571 18.45 1.83 Eastern 35.24 1.16 32.96 37.52 146,841 4,364,809 35.83 2.28 Khartoum 29.90 1.50 26.95 32.85 80,521 5,230,708 33.79 1.47 Central 27.05 0.85 25.39 28.72 129,942 7,295,131 29.10 1.08 Kordofan 44.23 1.07 42.13 46.33 164,120 4,229,456 43.66 3.95 Darfur 51.59 0.85 49.92 53.26 357,134 6,906,972 51.09 1.08 Source: Authors’ calculations based on the 2014/15 NHBPS and the 2008 Population and Housing Census. Looking at Table 14 and taking every step into account, one may infer that the methodology adopted to compute monetary poverty indicators at a lower level of spatial disaggregation is fully consistent with the poverty profile figures resulting from direct estimation from the survey. Therefore, standard errors of the poverty indicators can be computed, and the poverty maps are compatible with the poverty profile. It is probably a natural extension to the poverty profiles. 4. Poverty Mapping Results This section presents the results of the poverty mapping, tables and maps, in descending order. Poverty rate and Gini at national and regional levels are reported in Table 15. For the visualization, poverty by region is drawn in Map 1 while the number of poor by state is drawn in Map 2. Recall that the poverty mapping applies the results from the 2014/15 NHBPS to the census, the tables and maps below refer to these simulations, and the results are close to the survey. Table 15: Census SAE of Poverty and Gini at the National and Regional Levels Number Poverty Gini Head Standard. No. of Poor Standard Households Individuals Estimate Count Error Error Sudan 917,453 29,757,647 37.47 0.81 11,148,985 0.30 0.00 Northern 38,895 1,730,571 18.45 1.83 319,214 0.25 0.01 Eastern 146,841 4,364,809 35.83 2.28 1,563,727 0.27 0.02 Khartoum 80,521 5,230,708 33.79 1.47 1,767,652 0.30 0.01 Central 129,942 7,295,131 29.10 1.08 2,123,153 0.25 0.01 Kordofan 164,120 4,229,456 43.66 3.95 1,846,471 0.28 0.01 Darfur 357,134 6,906,972 51.09 1.08 3,528,766 0.38 0.00 Source: Authors’ calculations based on the 2014/15 NHBPS and the 2008 Population and Housing Census. 24 Results in Table 16 show that poverty rates within a region are not the same as for the region itself. They may exceed or be inferior; some regions may have lower poverty rates but states included in them may have higher poverty rates, or the opposite may hold true. The maximum difference noticed for all is about 10 percent. This is the case for the Central and Darfur regions. In the Central region, the poverty rate ranges from 21.24 percent in Al-Gezira state to 40.78 percent in White Nile state whereas the rate at the regional level is 29.10 percent. In the Darfur region, where at 51.09 percent, poverty rates are the highest, they range between 46.02 percent in East Darfur and 61.05 percent in Central Darfur. In addition to having the highest poverty rates, Darfur has the highest inequality (0.38). Consistently, its states’ inequality ranges from 0.365 to 0.388 while others do not exceed 0.304. Map 2 and Map 3 show the poverty rate at the state level and the number of poor for each state, respectively. Inequality statistics of Gini per state are shown in Map 4. At the lower level for each region, poverty rates results and Gini at the locality level are displayed in Table 17. It shows that poverty is a heterogenous phenomenon across counties. For instance, poverty rate in the Northern region is 18.45 percent with a standard error of 1.83 percent. However, at the locality level, the poverty rates range from 4.99 percent to 27.71 percent. Once again, this result justifies the choice of the disaggregated model. Another finding is that the regions that have the highest poverty rates at the regional and state levels tend to have lower poverty rates in the localities within them and vice versa. For example, Darfur with a poverty rate of 51.09 percent has a locality with a 33.02 percent poverty rate while Khartoum with a poverty rate of 33.79 percent has a locality with a 61.25 percent poverty rate. So poverty at the locality level is heterogenous and is not influenced by the poverty rate at the regional level. Map 5 displays the relative share of poor, Map 6 the number of poor, and Map 7 the Gini statistics by locality. Table 16: Census SAE of Poverty and Gini at the State Level Number Poverty Gini No. of Region State Poor Head Standard Standard Households Individuals Estimate Count Error Error Northern Northern 13,852 635,755 15.20 2.15 96,641 0.251 0.010 River Nile 25,043 1,094,816 20.33 2.41 222,574 0.255 0.009 Eastern Red Sea 62,682 1364398 42.16 4.21 575,229 0.259 0.015 Kassala 56,734 1683786 34.60 2.98 582,560 0.269 0.022 Al-Gedarif 27,425 1316625 30.83 3.30 405,938 0.262 0.017 Khartoum Khartoum 80,521 5,230,708 33.79 1.47 1,767,652 0.304 0.007 Central Al-Gezira 56,679 3,490,560 21.24 1.47 741,532 0.234 0.007 White 31,623 1,727,955 40.78 1.76 704,637 0.262 0.008 Nile 25 Number Poverty Gini No. of Region State Poor Head Standard Standard Households Individuals Estimate Count Error Error Sinnar 25,504 1,249,265 27.22 1.78 340,023 0.241 0.006 Blue Nile 16,136 827,349 40.73 2.39 336,961 0.249 0.008 Kordofan North 100,913 2,061,612 41.89 5.85 863,688 0.274 0.009 Kordofan South 23,025 866,698 53.19 5.89 461,006 0.275 0.009 Kordofan West 40,182 1,301,145 40.10 4.93 521,776 0.271 0.006 Kordofan Darfur North 88,597 2,118,507 50.54 2.12 1,070,630 0.378 0.005 Darfur West 26,942 790,383 58.99 1.15 466,258 0.388 0.005 Darfur South 144,580 2,541,122 49.64 1.39 1,261,450 0.374 0.004 Darfur Central 31,712 398,518 61.05 1.63 243,302 0.368 0.004 Darfur East 65,303 1,058,440 46.02 1.51 487,124 0.365 0.004 Darfur Source: Authors’ calculations based on the 2014/15 NHBPS and the 2008 Population and Housing Census. 26 Map 2: Census SAE of Poverty at the State Level Source: Based on the 2008 Population and Housing Census. 27 Map 3: Census SAE of Number of Poor at the State Level Source: Based on the 2008 Population and Housing Census. 28 Map 4: Census SAE of Gini at the State Level Source: Based on the 2008 Population and Housing Census. 29 Table 17: Census SAE of Poverty and Gini at the Locality Level Number Poverty Gini Locality No. of Region State Standard Standard Code Households Individuals Head Count Poor Estimate Error Error Northern 1111101 583 27,349 12.30 4.92 3,363 0.229 0.011 Northern 1111102 1,015 45,003 4.99 2.40 2,246 0.250 0.014 Northern 1111103 1,140 76,247 11.54 3.71 8,799 0.225 0.008 Northern 1111104 2,589 145,185 16.68 4.42 24,215 0.247 0.009 Northern 1111105 1,999 96,396 16.98 4.74 16,367 0.243 0.009 Northern 1111106 2,460 88,388 22.01 5.12 19,451 0.249 0.008 Northern Northern 1111107 4,066 157,184 14.12 4.40 22,201 0.245 0.008 River Nile 1121201 3,388 139,950 13.07 3.93 18,298 0.253 0.009 River Nile 1121202 2,728 148,938 20.41 4.77 30,395 0.244 0.008 River Nile 1121203 2,305 107,939 24.28 4.61 26,205 0.248 0.009 River Nile 1121204 8,461 272,756 27.71 6.22 75,575 0.262 0.010 River Nile 1121205 5,059 250,880 16.45 4.07 41,280 0.243 0.009 River Nile 1121206 3,102 174,351 17.68 4.31 30,820 0.241 0.008 Red Sea 2212101 5,940 91,773 37.71 11.07 34,604 0.234 0.014 Red Sea 2212102 4,581 172,700 36.86 9.41 63,650 0.239 0.039 Red Sea 2212103 7,673 409,912 44.62 9.05 182,899 0.272 0.012 Red Sea 2212104 6,121 93,366 32.11 8.69 29,981 0.268 0.024 Red Sea 2212105 11,813 112,928 49.91 9.31 56,363 0.255 0.019 Red Sea 2212106 9,339 249,351 41.93 9.89 104,542 0.239 0.013 Eastern Red Sea 2212107 10,649 146,469 35.12 9.25 51,435 0.233 0.014 Red Sea 2212108 6,566 87,895 58.88 11.25 51,753 0.224 0.023 Kassala 2222201 5,301 76,177 51.15 7.63 38,967 0.280 0.033 Kassala 2222202 6,703 178,316 37.48 11.15 66,829 0.205 0.019 Kassala 2222203 7,968 219,963 34.41 8.10 75,689 0.234 0.032 Kassala 2222204 3,454 94,410 30.69 8.19 28,972 0.258 0.019 Kassala 2222205 8,200 82,275 12.34 5.83 10,156 0.248 0.020 30 Number Poverty Gini Locality No. of Region State Standard Standard Code Households Individuals Head Count Poor Estimate Error Error Kassala 2222206 4,424 236,888 38.15 8.19 90,377 0.290 0.019 Kassala 2222207 8,426 147,832 41.64 9.34 61,554 0.247 0.018 Kassala 2222208 4,394 259,418 33.48 8.18 86,841 0.276 0.012 Kassala 2222209 4,500 191,272 25.63 8.31 49,024 0.251 0.042 Kassala 2222210 1,842 108,916 38.18 8.23 41,588 0.271 0.033 Kassala 2222211 1,522 88,315 36.87 8.77 32,561 0.263 0.037 Al-Gedarif 2232301 1,320 54,883 20.74 7.61 11,382 0.219 0.025 Al-Gedarif 2232302 3,332 133,859 20.93 7.65 28,018 0.240 0.016 Al-Gedarif 2232303 2,243 116,144 19.80 8.07 22,997 0.251 0.044 Al-Gedarif 2232304 4,368 259,148 47.66 10.07 123,511 0.255 0.013 Al-Gedarif 2232305 2,928 165,427 17.60 6.76 29,119 0.256 0.018 Al-Gedarif 2232306 4,530 215,084 23.88 7.46 51,368 0.239 0.019 Al-Gedarif 2232307 1,414 68,141 42.12 9.64 28,701 0.236 0.013 Al-Gedarif 2232308 2,439 90,673 33.25 8.57 30,148 0.252 0.020 Al-Gedarif 2232309 1,480 68,289 31.85 9.72 21,751 0.236 0.011 Al-Gedarif 2232310 3,371 144,973 40.66 9.04 58,940 0.249 0.021 Khartoum 3313101 11,084 720,130 38.58 3.19 277,853 0.275 0.006 Khartoum 3313102 16,763 1,117,428 61.25 2.94 684,444 0.268 0.008 Khartoum 3313103 6,599 418,831 20.95 3.53 87,763 0.291 0.008 Khartoum Khartoum 3313104 9,184 586,762 18.80 2.35 110,307 0.295 0.007 Khartoum 3313105 13,614 885,800 27.04 2.58 239,492 0.277 0.007 Khartoum 3313106 9,677 618,773 12.35 2.23 76,397 0.288 0.008 Khartoum 3313107 13,600 882,980 33.00 3.34 291,394 0.267 0.007 Al-Gezira 4414101 7,796 467,042 19.52 2.63 91,182 0.234 0.007 Al-Gezira 4414102 6,816 418,471 17.26 3.18 72,233 0.231 0.007 Central Al-Gezira 4414103 9,654 589,382 24.52 2.68 144,504 0.232 0.007 Al-Gezira 4414104 3,801 240,626 29.73 4.26 71,543 0.229 0.007 31 Number Poverty Gini Locality No. of Region State Standard Standard Code Households Individuals Head Count Poor Estimate Error Error Al-Gezira 4414105 6,305 381,603 22.08 3.13 84,272 0.251 0.011 Al-Gezira 4414106 8,901 564,915 20.16 3.40 113,863 0.223 0.006 Al-Gezira 4414107 13,406 828,516 19.79 2.59 163,933 0.231 0.007 White Nile 4424201 4,427 280,111 34.07 4.03 95,422 0.262 0.010 White Nile 4424202 1,518 91,756 28.20 3.28 25,878 0.255 0.007 White Nile 4424203 4,213 265,367 34.90 3.67 92,619 0.262 0.010 White Nile 4424204 3,337 239,867 39.07 4.47 93,709 0.260 0.010 White Nile 4424205 3,536 141,470 32.31 4.12 45,715 0.245 0.007 White Nile 4424206 6,751 422,600 47.82 3.61 202,070 0.258 0.010 White Nile 4424207 4,611 92,369 50.15 5.64 46,327 0.242 0.009 White Nile 4424208 3,230 194,413 52.93 4.90 102,897 0.246 0.008 Sinnar 4434301 3,337 200,056 25.86 4.08 51,732 0.241 0.007 Sinnar 4434302 5,095 289,581 31.76 3.56 91,984 0.251 0.007 Sinnar 4434303 5,199 167,890 28.39 3.51 47,660 0.229 0.009 Sinnar 4434304 3,891 245,534 29.45 3.23 72,299 0.233 0.006 Sinnar 4434305 2,430 147,961 24.31 4.47 35,970 0.249 0.009 Sinnar 4434306 2,426 122,156 19.16 4.26 23,410 0.223 0.009 Sinnar 4434307 3,126 76,084 22.30 5.75 16,968 0.233 0.011 Blue Nile 4444401 4,496 237,687 44.48 2.84 105,714 0.247 0.009 Blue Nile 4444402 3,378 212,765 42.08 2.80 89,525 0.257 0.009 Blue Nile 4444403 1,164 64,235 25.89 3.53 16,627 0.229 0.010 Blue Nile 4444404 2,872 132,062 48.78 8.51 64,423 0.223 0.010 Blue Nile 4444405 2,417 94,419 23.73 4.93 22,405 0.241 0.010 Blue Nile 4444406 1,809 86,178 44.40 8.00 38,266 0.240 0.012 North Kordofan 5515101 31,709 209,773 42.81 10.65 89,810 0.255 0.003 Kordofan North Kordofan 5515102 34,984 324,477 47.20 10.79 153,142 0.255 0.003 North Kordofan 5515103 8,115 387,736 42.71 10.23 165,599 0.274 0.004 32 Number Poverty Gini Locality No. of Region State Standard Standard Code Households Individuals Head Count Poor Estimate Error Error North Kordofan 5515104 13,612 612,660 41.56 12.24 254,600 0.271 0.004 North Kordofan 5515106 12,493 526,964 38.05 10.09 200,536 0.268 0.004 South Kordofan 5525201 5,857 204,056 48.54 10.05 99,058 0.270 0.005 South Kordofan 5525202 4,780 196,087 57.11 12.38 111,979 0.260 0.005 South Kordofan 5525203 5,462 271,372 48.64 10.59 131,992 0.271 0.005 South Kordofan 5525204 3,270 93,272 56.64 10.03 52,834 0.265 0.006 South Kordofan 5525206 3,656 101,908 63.92 10.13 65,141 0.255 0.005 West Kordofan 5535302 4,715 249,095 37.79 9.96 94,141 0.264 0.004 West Kordofan 5535303 2,655 126,062 39.01 10.63 49,173 0.263 0.005 West Kordofan 5535304 3,294 136,878 42.02 11.12 57,513 0.266 0.005 West Kordofan 5535307 6,537 284,177 36.78 10.20 104,533 0.265 0.004 West Kordofan 5535308 6,638 148,177 44.61 11.39 66,101 0.266 0.005 West Kordofan 5535309 3,906 128,074 40.68 11.09 52,099 0.266 0.005 West Kordofan 5535311 10,705 171,668 42.93 10.30 73,695 0.259 0.004 West Kordofan 5535312 1,732 57,010 43.01 9.92 24,519 0.265 0.007 North Darfur 6616101 6,878 166,801 60.37 3.88 100,698 0.349 0.008 North Darfur 6616102 1,058 162,038 48.98 5.70 79,359 0.356 0.008 North Darfur 6616104 1,396 224,876 59.63 8.86 134,083 0.357 0.006 North Darfur 6616105 10,680 201,532 57.85 9.68 116,580 0.363 0.007 North Darfur 6616106 9,800 139,054 44.98 3.55 62,553 0.367 0.007 North Darfur 6616107 1,079 100,743 56.87 4.15 57,289 0.358 0.009 Darfur North Darfur 6616108 4,325 101,450 55.12 7.11 55,924 0.341 0.012 North Darfur 6616109 3,461 300,105 44.69 2.74 134,114 0.398 0.009 North Darfur 6616110 1,813 166,843 38.88 4.13 64,870 0.358 0.008 North Darfur 6616111 819 118,086 59.34 6.28 70,072 0.359 0.011 North Darfur 6616112 1,596 203,375 33.02 4.93 67,148 0.360 0.007 North Darfur 6616114 45,692 233,598 54.77 10.55 127,937 0.362 0.003 33 Number Poverty Gini Locality No. of Region State Standard Standard Code Households Individuals Head Count Poor Estimate Error Error West Darfur 6626201 5,314 90,416 73.60 2.54 66,543 0.354 0.009 West Darfur 6626202 4,231 114,074 75.27 2.14 85,868 0.364 0.007 West Darfur 6626203 4,266 97,936 44.23 3.34 43,320 0.362 0.007 West Darfur 6626204 3,364 245,568 49.01 2.37 120,349 0.388 0.007 West Darfur 6626205 4,179 91,438 68.17 2.93 62,333 0.364 0.007 West Darfur 6626206 5,588 150,947 58.19 2.94 87,842 0.364 0.006 South Darfur 6636302 11,539 428,808 46.69 2.68 200,196 0.391 0.006 South Darfur 6636303 2,251 11,994 60.47 9.30 7,253 0.358 0.005 South Darfur 6636304 10,110 156,341 56.08 3.45 87,681 0.364 0.005 South Darfur 6636305 48,384 709,985 54.33 2.95 385,748 0.359 0.003 South Darfur 6636306 8,456 243,551 56.13 3.90 136,717 0.364 0.005 South Darfur 6636309 37,243 400,784 34.25 3.60 137,287 0.354 0.004 South Darfur 6636310 16,075 274,755 64.62 3.37 177,553 0.360 0.004 South Darfur 6636311 10,522 314,900 40.97 3.22 129,014 0.350 0.004 Central Darfur 6646407 5,280 119,475 62.94 3.22 75,202 0.363 0.006 Central Darfur 6646408 5,887 32,245 66.31 2.17 21,381 0.365 0.006 Central Darfur 6646409 2,046 11,265 67.68 3.89 7,624 0.362 0.006 Central Darfur 6646411 10,289 175,870 57.56 2.49 101,227 0.361 0.005 Central Darfur 6646412 4,877 32,119 76.58 2.08 24,597 0.364 0.008 Central Darfur 6646413 3,333 27,540 48.18 3.18 13,268 0.362 0.011 East Darfur 6656501 11,481 182,629 50.92 3.07 92,998 0.368 0.005 East Darfur 6656507 18,355 447,965 50.49 1.99 226,199 0.364 0.005 East Darfur 6656508 4,594 202,180 36.33 3.17 73,448 0.354 0.007 East Darfur 6656512 30,873 225,664 41.87 3.33 94,478 0.358 0.004 Source: Authors’ calculations based on the 2014/15 NHBPS and the 2008 Population and Housing Census. 34 Map 5: Census SAE of Poverty at the Locality Level Source: Based on the 2008 Population and Housing Census. Note: No data imply that the locality was not in the Sample Census 2008. 35 Map 6: Census SAE of Number of Poor at the Locality Level Source: Based on the 2008 Population and Housing Census. Note: No data imply that the locality was not in the Sample Census 2008. 36 Map 7: Census SAE Gini at the Locality Level Source: Based on the 2008 Population and Housing Census. Note: No data imply that the locality was not in the Sample Census 2008. 37 5. Conclusions Poverty maps, unlike household surveys, allow for a greater focus on the spatial distribution of poverty and inequality at lower levels. This analysis highlights the potential gains from a more aggregate level to a lower-level geographical targeting. This may offer an effective approach for reaching the poor where there are substantial disparities in living standards within and across geographical areas. Although it is beyond the scope of this note, this methodology can be complemented with other indicators of well-being, opportunity, and access for regional patterns. This report describes the methodology of the small area poverty and inequality estimation presenting the results for the 131 localities of Sudan using the 2014/15 NHBPS and the 2008 Population and Housing Census. Variables such as demographics characteristics, education, occupation, housing characteristics, and productive and durable assets are used in the modeling. Since the Sudan poverty phenomenon is heterogenous, the disaggregated model is seen as a better fit compared to the national model. Estimates of poverty head count at the national and state levels are found to be in line with the poverty levels observed in the 2014/15 NHBPS. Northern region has the lowest poverty head count and Darfur region has the highest poverty head count as well as inequality. Overall inequality is moderate. Poverty rates at state or locality level do not always correspond with data at regional levels. These differences are induced by the heterogenous character of the poverty phenomenon. The poverty and inequality maps presented in this report may find use for other future analysis. Poverty estimates for localities may be used as inputs to conduct the analysis at the district level or to look at specific non-geographical factors associated with poverty and inequality. These results may also provide decision-makers with a starting point for improving the targeting of poverty reduction strategies. 38 References Ahmed, Faizuddin, C. Cheku, S. Takamatsu, and N. Yoshida. 2014. Hybrid Survey to Improve the Reliability of Poverty Statistics in a Cost-Effective Manner. Bedi, T., A. Coudouel, and K. Simler, eds. 2007. More Than a Pretty Picture: Using Poverty Maps to Design Better Policies and Interventions. Washington, DC: World Bank. Bigman, D., and U. Deichmann. 2000. “Geographic Targeting: A Review of Different Approaches.” In Geographical Targeting for Poverty Alleviation: Methodology and Applications, edited by D. Bigman and H. Fofack, 43–73. Washington, DC: World Bank. CBS (Central Bureau of Statistics). 2017. “Sudan: Poverty Profile in 2014 - Main Findings.” Khartoum: CBS. Coudouel, A., J. S. Hentschel, and Q. T. Wodon. 2002. “Poverty Measurement and Analysis.” In A Sourcebook for Poverty Reduction Strategies: Core Techniques and Cross-Cutting Issues, edited by J. Klugman, 27–74. Washington, DC: World Bank Group. Elbers, C., J. O. Lanjouw, and P. Lanjouw. 2003. “Micro-level Estimation of Poverty and Inequality.” Econometrica 71: 355–364. Foster, James E., J. Greer, E. Thorbecke. 1984. “A Class of Decomposable Poverty Indices.” Econometrica 52 (3): 761–766. Henderson, C. R. 1953. “Estimation of Variance and Covariance Components.” Biometrics 9: 226– 252. James, Arthur, Michael Waring, Robert Coe, Larry V. Hedges eds. 2012. “Research Methods and Methodologies in Education.” British Journal of Educational Technology 44 (2): E63–E64. Lindsey, C., and S. J. Sheather. 2010. “Variable Selection in Linear Regression.” Stata Journal 10 (4): 650–669. Mungai, Rose, Minh Cong Nguyen, and Tejesh Pradhan. 2018. “Poverty and Inequality on the Map in The Gambia: An Application of Small Area Estimation.” World Bank: Washington, DC. Searle, Shayle R., G. Casella, and C. E. McCulloch. 1992. Variance Components. Wiley, New York. Thompson, Bruce. 1995. “Stepwise Regression and Stepwise Discriminant Analysis Need Not Apply Here: A Guidelines Editorial.” Educational and Psychological Measurement 55: 525– 534. UNFPA (United Nations Population Fund). 2017. 39 40 Appendix A: Sudan Administrative Boundaries Source: www.mapsofworld.com/July 2014. 41 Appendix B: Common Variables between the Census and 2014/15 NHBPS Individual Information Age Gender 0 Female 1 Male Relationship to the head of HH 1 Head 2 Spouse 3 Child 4 Parents 5 Other relative 6 Non-relative Marital status 1 Never married 2 Married 3 Widowed 4 Divorced Able to read or write 0 Illiterate 1 Literate Has ever attended school 0 not attended 1 Attended Is currently attending school 0 not attending 1 Attending Highest education level attainted 1 None 2 Incomplete primary 3 Primary 4 Secondary 5 Tertiary Level of schooling currently attending 1 Primary 2 Secondary 3 Tertiary Employed 0 No 1 Yes 42 Individual Information why not in labor force 1 No hope to find job 2 Full time student 3 Income recipient 4 Too old 5 Disabled/too sick Full time 6 homemaker/housewife 7 Pensioner/retired type of employment 1 Paid employee 2 Employer 3 Own account worker 4 Unpaid family worker 5 Unpaid working for others Industry (Main industry) Sector of work 1 Agr 2 Mnf 3 Services Any disability? Difficulty in seeing 0 No 1 Yes Blindness 0 No 1 Yes Difficulty in hearing 0 No 1 Yes Deafness 0 No 1 Yes Difficulty in speaking 0 No 1 Yes Mutism 0 No 1 Yes Disability in other part of the body 0 No 43 Individual Information 1 Yes Mental retardation 0 No 1 Yes Other disability type 0 No 1 Yes Father alive? 1 Dead 2 Alive 3 Not know Mother alive? 1 Dead 2 Alive 3 Not know 44 Household Information Household size Age (head) Number of rooms used for sleeping Number of livestock and poultry 1 Cattle 2 Horses 3 Donkeys 4 Sheep 5 Goats 6 Poultry 7 Camels Region 1 Northern 2 Eastern 3 Khartoum 4 Central 5 Kordofan 6 Darfur Rural or Urban 0 Rural 1 Urban Type of dwelling 1 Tent 2 Dwelling of straw mats 3 Gottiya-mud 4 Gottiya-sticks 5 Apartment 6 Villa 7 House of one floor-mud 8 House of one floor-brick/concrete 9 House constructed of wood 10 Multi-storey house 11 Incomplete Ownership of dwelling 1 Owned 2 Rented 3 Housing provided as part of work 4 Free Access to drinking water 45 Household Information 0 No 1 Yes Access to electricity 0 No 1 Yes Energy used for cooking 1 Firewood 2 Charcoal 3 Gas 4 Electricity 5 Paraffin 6 Cow dung 7 Grass 8 Biogas 9 No cooking Type of toilet 1 Pit latrine private 2 Shared pit latrine 3 Private flush toilet 4 Shate flush toilet 5 Bucket toilet 6 No toilet facility Assets tv 0 No 1 Yes Radio 0 No 1 Yes Phone 0 No 1 Yes computer 0 No 1 Yes refrigerator 0 No 1 Yes Fan 0 No 46 Household Information 1 Yes Ac 0 No 1 Yes motor 0 No 1 Yes motorcycle 0 No 1 Yes bicycle 0 No 1 Yes boat 0 No 1 Yes anitran 0 No 1 Yes agriland 0 No 1 Yes Land ownership 1 Owned 2 Rented 3 Partially owned 4 Communal Forms of livelihood 1 Crop farming 2 Animal husbandry 3 Wages and salaries 4 Owned business enterprise 5 Property income 6 Remittances 7 Pension 8 Aid 9 Others 47 Appendix C: Region Alpha Model Estimates Table C.1: Northern (Alpha Model, Region 1) Coefficient Standard Error P >|z| head_age25t64 −135.941 68.789 0.048 head_edlevel4 9.496 4.281 0.027 hhsize_5 −12.953 10.308 0.209 dwelling1_yhat 0.115 0.046 0.013 head_age25t64_yhat 29.175 15.094 0.053 head_edlevel4_yhat −1.029 0.474 0.030 hhsize_4_yhat −0.059 0.026 0.022 hhsize_5_yhat 1.358 1.145 0.236 hhsize_6_yhat −0.037 0.027 0.170 nrooms_yhat −0.028 0.010 0.003 sector1_share_yhat −0.094 0.040 0.018 head_age25t64_yhat2 −1.566 0.827 0.058 Constant −4.033 0.317 0.000 Table C.2: Eastern (Alpha Model, Region 2) Coefficient Standard Error P >|z| charcoal 165.980 78.269 0.034 child_2 66.230 90.890 0.466 dwelling3 −105.706 63.141 0.094 elec_m_county −285.426 138.722 0.040 Gas 232.837 93.291 0.013 head_edlevel2 −170.842 92.598 0.065 head_edlevel3 489.506 227.036 0.031 head_unpaid 519.549 462.419 0.261 hhsize_2 −447.455 251.435 0.075 nrooms 3.756 1.860 0.043 sum_age1t14_m_county −5.807 37.175 0.876 charcoal_yhat −37.844 17.725 0.033 child_2_yhat −15.245 20.412 0.455 dwelling3_yhat 23.313 14.110 0.098 elec_m_county_yhat 63.621 31.328 0.042 gas_yhat −53.392 20.950 0.011 head_edlevel2_yhat 37.034 20.539 0.071 head_edlevel3_yhat −106.442 49.409 0.031 head_employed_yhat -0.056 0.022 0.011 head_employer_yhat 0.041 0.024 0.096 48 Coefficient Standard Error P >|z| head_unpaid_yhat −115.306 101.985 0.258 hhsize_2_yhat 97.062 53.554 0.070 hhsize_6_yhat 0.040 0.019 0.041 nrooms_yhat −0.428 0.209 0.040 sum_age1t14_m_county_yhat 1.542 8.309 0.853 charcoal_yhat2 2.154 1.003 0.032 child_2_yhat2 0.872 1.146 0.447 dwelling3_yhat2 −1.285 0.787 0.103 elec_m_county_yhat2 −3.524 1.767 0.046 gas_yhat2 3.052 1.175 0.009 head_edlevel2_yhat2 −2.001 1.138 0.079 head_edlevel3_yhat2 5.774 2.685 0.032 head_unpaid_yhat2 6.381 5.614 0.256 hhsize_2_yhat2 −5.259 2.852 0.065 sector3_share_yhat2 0.005 0.003 0.058 sum_age1t14_m_county_yhat2 −0.097 0.464 0.834 toilet2_yhat2 0.007 0.003 0.025 Constant −4.988 0.548 0.000 Table C.3: Khartoum (Alpha model, Region 3) Coefficient Standard Error P >|z| hhsize_6 9.247 4.794 0.054 everattend_share_yhat −0.185 0.080 0.020 hhsize_6_yhat −1.035 0.540 0.055 pri_abv_share_yhat −0.073 0.032 0.024 sec_abv_share_yhat 0.105 0.032 0.001 Constant −3.234 0.683 0.000 Table C.4: Central (Alpha model, Region 4) Coefficient Standard Error P >|z| dwelling3_m_county 6.706 3.326 0.044 head_employee 84.187 39.203 0.032 hhsize_4 −0.470 0.158 0.003 nrooms −2.621 1.247 0.036 toilet1 −0.366 0.117 0.002 urban 0.231 0.117 0.048 water_m_state −1.381 0.435 0.001 dwelling3_m_county_yhat −0.678 0.376 0.071 head_employee_yhat −18.761 8.757 0.032 49 Coefficient Standard Error P >|z| nrooms_yhat 0.294 0.140 0.035 head_employee_yhat2 1.041 0.489 0.033 pri_abv_share_yhat2 −0.004 0.002 0.037 Constant -4.549 0.212 0.000 50 Appendix D: Poverty Measures This section provides the mathematical expressions for the poverty measures used in the paper and for the World Bank. Three poverty measures of the Foster-Greer-Thorbecke (FGT) class (Foster, Greer, and Thorbecke 1984) are used—the head count, the poverty gap, and the squared poverty gap. For a simple introduction to poverty measurement and profiles, see Coudouel Hentschel, and Wodon (2002). The poverty head count is the share of the population which is poor, that is, the proportion of the population for whom consumption per equivalent adult y is less than the poverty line z. If we consider a population of size n in which q people are poor, then the head count index is defined as q H = n . The poverty gap, which is often considered as representing the depth of poverty, is the mean distance separating the population from the poverty line, with the nonpoor being given a distance of zero. Arranging consumption in ascending order y1,...., yq < z < yq+1, ..., yn with the poorest household’s consumption denoted by y1, the next poorest y2, and so on, and the richest household’s consumption by yn, the poverty gap is defined as follows: 1 q  z − yi  PG =  n i=1   z ,  where yi is the income of individual i, and the sum is taken only on those individuals who are poor, although in practice, we often work with household consumption rather than individual consumption. The poverty gap is thus a measure of the poverty deficit of the entire population where the notion of ‘poverty deficit’ captures resources that would be needed—as a proportion of the poverty line—to lift all the poor out of poverty through perfectly targeted cash transfers. The squared poverty gap is often described as a measure of the severity of poverty. While the poverty gap considers the distance separating the poor from the poverty line, the squared poverty gap takes the square of that distance into account. When using the squared poverty gap, the poverty gap is weighted by itself, so as to give more weight to the very poor. In other words, the squared poverty gap takes into account the inequality among the poor. It is defined as follows: 2 1 q  z − yi  SPG =   n i =1  z   . 51 The head count, the poverty gap, and the squared poverty gap are the first three measures of the FGT class of poverty measures and a common structure is evident that suggests a generic class of additive measures. It must be noted that the additive measures are such that aggregate poverty is equal to the population-weighted sum of poverty in various subgroups of society. The general formula for the class of poverty measures depends on a parameter α which takes a value of 0 for the head count, 1 for the poverty gap, and 2 for the squared poverty gap in the following expression:  1 q  z − yi  P =    (  0) n i =1  z  . The discussion that follows focuses on the head count index of poverty. Higher-order poverty measures—poverty gap and squared poverty gap—are provided in Appendix E. 52 Appendix E: Census Poverty Measures by Administrative Units Table E.1: Poverty Measures by Region and State Severity Contribution of Poverty Head Poverty Population Number of Region State of Head Poverty Severity of Population Size Count Gap Share Poor Poverty Count Gap Poverty SUDAN 37.47 10.58 4.37 100 100 100 100 29,757,647 11,148,985 Northern 18.45 3.72 1.19 5.82 2.86 2.04 1.58 1,730,571 319,214 Northern 15.20 2.85 0.86 2.14 0.87 0.58 0.42 635,755 96,641 River Nile 20.33 4.22 1.38 3.68 2.00 1.47 1.16 1,094,816 222,574 Eastern 35.83 8.27 2.80 14.67 14.03 11.47 9.39 4,364,809 1,563,727 Red Sea 42.16 10.03 3.45 4.59 5.16 4.35 3.62 1,364,398 575,229 Kassala 34.60 7.90 2.65 5.66 5.23 4.23 3.43 1,683,786 582,560 Al-Gedarif 30.83 6.91 2.31 4.42 3.64 2.89 2.34 1,316,625 405,938 Khartoum Khartoum 33.79 8.51 3.07 17.58 15.85 14.14 12.37 5,230,708 1,767,652 Central 29.10 6.34 2.09 24.52 19.04 14.70 11.71 7,295,131 2,123,153 Al-Gezira 21.24 3.95 1.15 11.73 6.65 4.38 3.07 3,490,560 741,532 White Nile 40.78 10.31 3.75 5.81 6.32 5.66 4.98 1,727,955 704,637 Sinnar 27.22 5.54 1.73 4.20 3.05 2.20 1.66 1,249,265 340,023 Blue Nile 40.73 9.37 3.13 2.78 3.02 2.46 1.99 827,349 336,961 Kordofan 43.66 11.39 4.16 14.21 16.56 15.30 13.54 4,229,456 1,846,471 North Kordofan 41.89 10.76 3.89 6.93 7.75 7.05 6.16 2,061,612 863,688 South Kordofan 53.19 15.25 5.95 2.91 4.13 4.20 3.96 866,698 461,006 West Kordofan 40.10 9.82 3.41 4.37 4.68 4.06 3.41 1,301,145 521,776 Darfur 51.09 19.30 9.68 23.21 31.65 42.35 51.41 6,906,972 3,528,766 North Darfur 50.54 18.94 9.43 7.12 9.60 12.75 15.37 2,118,507 1,070,630 West Darfur 58.99 24.30 12.94 2.66 4.18 6.10 7.86 790,383 466,258 South Darfur 49.64 18.43 9.12 8.54 11.31 14.88 17.83 2,541,122 1,261,450 Central Darfur 61.05 24.86 13.08 1.34 2.18 3.15 4.01 398,518 243,302 East Darfur 46.02 16.29 7.80 3.56 4.37 5.48 6.34 1,058,440 487,124 Source: Authors’ calculations based on the 2014/15 NHBPS and the 2008 Population and Housing Census. 53 Table E.2: Poverty Measures by Locality Level Severity Contribution of Poverty Head Poverty Population Population Number of Locality of Head Poverty Severity of Region State Count Gap Share Size Poor Code Poverty Count Gap Poverty SUDAN 37.47 10.58 4.37 100 100 100 100 29,757,647.00 11,148,985.00 18.45 3.72 1.19 5.82 2.86 2.04 1.58 1,730,571.00 319,214.00 Northern 1111101 12.30 1.84 0.44 0.09 0.03 0.02 0.01 27,349.00 3,363.00 Northern 1111102 4.99 0.75 0.19 0.15 0.02 0.01 0.01 45,003.00 2,245.89 Northern 1111103 11.54 1.91 0.51 0.26 0.08 0.05 0.03 76,247.00 8,798.65 Northern 1111104 16.68 3.06 0.89 0.49 0.22 0.14 0.10 145,185.00 24,214.59 Northern 1111105 16.98 3.21 0.98 0.32 0.15 0.10 0.07 96,396.00 16,367.13 Northern 1111106 22.01 4.62 1.52 0.30 0.17 0.13 0.10 88,388.00 19,450.83 Northern Northern 1111107 14.12 2.69 0.83 0.53 0.20 0.13 0.10 157,184.00 22,200.59 River Nile 1121201 13.07 2.45 0.75 0.47 0.16 0.11 0.08 139,950.00 18,297.63 River Nile 1121202 20.41 4.16 1.36 0.50 0.27 0.20 0.16 148,938.00 30,395.48 River Nile 1121203 24.28 4.87 1.52 0.36 0.24 0.17 0.13 107,939.00 26,205.36 River Nile 1121204 27.71 6.37 2.25 0.92 0.68 0.55 0.47 272,756.00 75,574.92 River Nile 1121205 16.45 3.13 0.95 0.84 0.37 0.25 0.18 250,880.00 41,280.11 River Nile 1121206 17.68 3.49 1.09 0.59 0.28 0.19 0.15 174,351.00 30,819.64 35.83 8.27 2.80 14.67 14.03 11.47 9.39 4,364,809.00 1,563,727.00 Red Sea 2212101 37.71 8.06 2.55 0.31 0.31 0.23 0.18 91,773.00 34,604.00 Red Sea 2212102 36.86 8.34 2.84 0.58 0.57 0.46 0.38 172,700.00 63,650.00 Red Sea 2212103 44.62 11.00 3.88 1.38 1.64 1.43 1.22 409,912.00 182,899.00 Red Sea 2212104 32.11 7.29 2.44 0.31 0.27 0.22 0.18 93,366.00 29,981.21 Eastern Red Sea 2212105 49.91 13.33 5.02 0.38 0.51 0.48 0.44 112,928.00 56,362.58 Red Sea 2212106 41.93 9.38 3.02 0.84 0.94 0.74 0.58 249,351.00 104,541.86 Red Sea 2212107 35.12 7.77 2.53 0.49 0.46 0.36 0.29 146,469.00 51,434.85 Red Sea 2212108 58.88 15.22 5.41 0.30 0.46 0.42 0.37 87,895.00 51,752.72 Kassala 2222201 51.15 13.65 5.06 0.26 0.35 0.33 0.30 76,177.00 38,967.39 54 Level Severity Contribution of Poverty Head Poverty Population Population Number of Locality of Head Poverty Severity of Region State Count Gap Share Size Poor Code Poverty Count Gap Poverty Kassala 2222202 37.48 7.82 2.42 0.60 0.60 0.44 0.33 178,316.00 66,829.37 Kassala 2222203 34.41 7.55 2.49 0.74 0.68 0.53 0.42 219,963.00 75,688.67 Kassala 2222204 30.69 6.76 2.21 0.32 0.26 0.20 0.16 94,410.00 28,972.09 Kassala 2222205 12.34 2.10 0.57 0.28 0.09 0.06 0.04 82,275.00 10,155.94 Kassala 2222206 38.15 9.33 3.28 0.80 0.81 0.70 0.60 236,888.00 90,376.99 Kassala 2222207 41.64 9.79 3.29 0.50 0.55 0.46 0.37 147,832.00 61,553.76 Kassala 2222208 33.48 7.64 2.52 0.87 0.78 0.63 0.50 259,418.00 86,841.17 Kassala 2222209 25.63 5.12 1.57 0.64 0.44 0.31 0.23 191,272.00 49,023.64 Kassala 2222210 38.18 9.28 3.27 0.37 0.37 0.32 0.27 108,916.00 41,587.89 Kassala 2222211 36.87 8.75 3.03 0.30 0.29 0.25 0.21 88,315.00 32,560.81 Al-Gedarif 2232301 20.74 3.54 0.95 0.18 0.10 0.06 0.04 54,883.00 11,382.49 Al-Gedarif 2232302 20.93 3.81 1.08 0.45 0.25 0.16 0.11 133,859.00 28,017.71 Al-Gedarif 2232303 19.80 4.02 1.28 0.39 0.21 0.15 0.11 116,144.00 22,997.06 Al-Gedarif 2232304 47.66 12.34 4.50 0.87 1.11 1.02 0.90 259,148.00 123,511.50 Al-Gedarif 2232305 17.60 3.38 1.01 0.56 0.26 0.18 0.13 165,427.00 29,119.06 Al-Gedarif 2232306 23.88 4.61 1.38 0.72 0.46 0.32 0.23 215,084.00 51,367.86 Al-Gedarif 2232307 42.12 9.82 3.34 0.23 0.26 0.21 0.17 68,141.00 28,700.73 Al-Gedarif 2232308 33.25 7.31 2.37 0.30 0.27 0.21 0.17 90,673.00 30,148.35 Al-Gedarif 2232309 31.85 6.32 1.87 0.23 0.20 0.14 0.10 68,289.00 21,750.81 Al-Gedarif 2232310 40.66 9.73 3.40 0.49 0.53 0.45 0.38 144,973.00 58,939.97 33.79 8.51 3.07 17.58 15.85 14.14 12.37 5,230,708 1,767,652 Khartoum 3313101 38.58 9.25 3.18 2.42 2.49 2.12 1.76 720,130.00 277,852.98 Khartoum 3313102 61.25 18.53 7.46 3.76 6.14 6.58 6.41 1,117,428.00 684,444.43 Khartoum Khartoum 3313103 20.95 4.23 1.29 1.41 0.79 0.56 0.41 418,831.00 87,763.41 Khartoum 3313104 18.80 3.75 1.13 1.97 0.99 0.70 0.51 586,762.00 110,306.51 Khartoum 3313105 27.04 5.76 1.83 2.98 2.15 1.62 1.24 885,800.00 239,492.09 Khartoum 3313106 12.35 2.25 0.64 2.08 0.69 0.44 0.30 618,773.00 76,397.22 55 Level Severity Contribution of Poverty Head Poverty Population Population Number of Locality of Head Poverty Severity of Region State Count Gap Share Size Poor Code Poverty Count Gap Poverty Khartoum 3313107 33.00 7.56 2.53 2.97 2.61 2.12 1.72 882,980.00 291,394.14 29.10 6.34 2.09 24.52 19.04 14.70 11.71 7,295,131 2,123,153 Al-Gezira 4414101 19.52 3.44 0.95 1.57 0.82 0.51 0.34 467,042.00 91,181.58 Al-Gezira 4414102 17.26 3.02 0.84 1.41 0.65 0.40 0.27 418,471.00 72,232.90 Al-Gezira 4414103 24.52 4.69 1.39 1.98 1.30 0.88 0.63 589,382.00 144,504.41 Al-Gezira 4414104 29.73 5.97 1.82 0.81 0.64 0.46 0.34 240,626.00 71,542.93 Al-Gezira 4414105 22.08 4.45 1.40 1.28 0.76 0.54 0.41 381,603.00 84,272.13 Al-Gezira 4414106 20.16 3.60 1.00 1.90 1.02 0.65 0.44 564,915.00 113,863.49 Al-Gezira 4414107 19.79 3.58 1.02 2.78 1.47 0.94 0.65 828,516.00 163,933.41 White Nile 4424201 34.07 8.17 2.89 0.94 0.86 0.73 0.62 280,111.00 95,421.59 White Nile 4424202 28.20 6.18 2.03 0.31 0.23 0.18 0.14 91,756.00 25,878.10 White Nile 4424203 34.90 8.44 2.98 0.89 0.83 0.71 0.61 265,367.00 92,618.52 White Nile 4424204 39.07 9.86 3.63 0.81 0.84 0.75 0.67 239,867.00 93,708.59 White Nile 4424205 32.31 7.04 2.27 0.48 0.41 0.32 0.25 141,470.00 45,714.96 Central White Nile 4424206 47.82 13.00 5.01 1.42 1.81 1.75 1.63 422,600.00 202,069.84 White Nile 4424207 50.15 12.69 4.47 0.31 0.42 0.37 0.32 92,369.00 46,327.17 White Nile 4424208 52.93 13.84 5.01 0.65 0.92 0.85 0.75 194,413.00 102,897.14 Sinnar 4434301 25.86 5.26 1.63 0.67 0.46 0.33 0.25 200,056.00 51,731.72 Sinnar 4434302 31.76 7.17 2.42 0.97 0.83 0.66 0.54 289,581.00 91,983.68 Sinnar 4434303 28.39 5.40 1.56 0.56 0.43 0.29 0.20 167,890.00 47,660.37 Sinnar 4434304 29.45 5.98 1.86 0.83 0.65 0.47 0.35 245,534.00 72,299.39 Sinnar 4434305 24.31 5.01 1.58 0.50 0.32 0.24 0.18 147,961.00 35,969.68 Sinnar 4434306 19.16 3.24 0.86 0.41 0.21 0.13 0.08 122,156.00 23,409.58 Sinnar 4434307 22.30 3.75 0.98 0.26 0.15 0.09 0.06 76,084.00 16,967.56 Blue Nile 4444401 44.48 10.57 3.62 0.80 0.95 0.80 0.66 237,687.00 105,713.60 Blue Nile 4444402 42.08 10.40 3.71 0.71 0.80 0.70 0.61 212,765.00 89,524.89 Blue Nile 4444403 25.89 4.65 1.28 0.22 0.15 0.09 0.06 64,235.00 16,627.25 56 Level Severity Contribution of Poverty Head Poverty Population Population Number of Locality of Head Poverty Severity of Region State Count Gap Share Size Poor Code Poverty Count Gap Poverty Blue Nile 4444404 48.78 10.84 3.43 0.44 0.58 0.45 0.35 132,062.00 64,423.37 Blue Nile 4444405 23.73 4.31 1.20 0.32 0.20 0.13 0.09 94,419.00 22,405.14 Blue Nile 4444406 44.40 10.27 3.42 0.29 0.34 0.28 0.23 86,178.00 38,265.99 43.66 11.39 4.16 14.21 16.56 15.30 13.54 4,229,456 1,846,471 North Kordofan 5515101 42.81 10.79 3.84 0.70 0.81 0.72 0.62 209,773.00 89,810.06 North Kordofan 5515102 47.20 12.33 4.48 1.09 1.37 1.27 1.12 324,477.00 153,141.93 North Kordofan 5515103 42.71 11.13 4.06 1.30 1.49 1.37 1.21 387,736.00 165,599.35 North Kordofan 5515104 41.56 10.79 3.93 2.06 2.28 2.10 1.85 612,660.00 254,600.24 North Kordofan 5515106 38.05 9.47 3.36 1.77 1.80 1.59 1.36 526,964.00 200,535.93 South Kordofan 5525201 48.54 13.18 4.94 0.69 0.89 0.85 0.77 204,056.00 99,058.21 South Kordofan 5525202 57.11 16.84 6.69 0.66 1.00 1.05 1.01 196,087.00 111,979.26 South Kordofan 5525203 48.64 13.33 5.04 0.91 1.18 1.15 1.05 271,372.00 131,992.36 Kordofan South Kordofan 5525204 56.64 16.88 6.75 0.31 0.47 0.50 0.48 93,272.00 52,833.82 South Kordofan 5525206 63.92 19.96 8.23 0.34 0.58 0.65 0.65 101,908.00 65,140.54 West Kordofan 5535302 37.79 8.99 3.06 0.84 0.84 0.71 0.59 249,095.00 94,141.01 West Kordofan 5535303 39.01 9.45 3.25 0.42 0.44 0.38 0.32 126,062.00 49,173.42 West Kordofan 5535304 42.02 10.61 3.79 0.46 0.52 0.46 0.40 136,878.00 57,512.76 West Kordofan 5535307 36.78 8.64 2.91 0.95 0.94 0.78 0.64 284,177.00 104,533.45 West Kordofan 5535308 44.61 11.38 4.06 0.50 0.59 0.54 0.46 148,177.00 66,101.22 West Kordofan 5535309 40.68 10.20 3.62 0.43 0.47 0.42 0.36 128,074.00 52,099.23 West Kordofan 5535311 42.93 10.64 3.72 0.58 0.66 0.58 0.49 171,668.00 73,694.57 West Kordofan 5535312 43.01 10.82 3.82 0.19 0.22 0.20 0.17 57,010.00 24,518.72 51.09 19.30 9.68 23.21 31.65 42.35 51.41 6,906,972 3,528,766 North Darfur 6616101 60.37 23.47 11.88 0.56 0.90 1.24 1.52 166,801.00 100,697.81 Darfur North Darfur 6616102 48.98 17.46 8.37 0.54 0.71 0.90 1.04 162,038.00 79,359.32 North Darfur 6616104 59.63 23.70 12.25 0.76 1.20 1.69 2.12 224,876.00 134,083.15 57 Level Severity Contribution of Poverty Head Poverty Population Population Number of Locality of Head Poverty Severity of Region State Count Gap Share Size Poor Code Poverty Count Gap Poverty North Darfur 6616105 57.85 23.10 12.01 0.68 1.05 1.48 1.86 201,532.00 116,579.51 North Darfur 6616106 44.98 15.95 7.65 0.47 0.56 0.70 0.82 139,054.00 62,552.82 North Darfur 6616107 56.87 22.05 11.21 0.34 0.51 0.71 0.87 100,743.00 57,289.42 North Darfur 6616108 55.12 20.44 10.02 0.34 0.50 0.66 0.78 101,450.00 55,923.73 North Darfur 6616109 44.69 16.38 8.05 1.01 1.20 1.56 1.86 300,105.00 134,114.45 North Darfur 6616110 38.88 12.70 5.74 0.56 0.58 0.67 0.74 166,843.00 64,869.86 North Darfur 6616111 59.34 23.41 12.00 0.40 0.63 0.88 1.09 118,086.00 70,071.94 North Darfur 6616112 33.02 10.35 4.53 0.68 0.60 0.67 0.71 203,375.00 67,148.26 North Darfur 6616114 54.77 21.33 10.91 0.79 1.15 1.58 1.96 233,598.00 127,936.96 West Darfur 6626201 73.60 33.21 18.61 0.30 0.60 0.95 1.29 90,416.00 66,543.45 West Darfur 6626202 75.27 35.23 20.28 0.38 0.77 1.28 1.78 114,074.00 85,867.87 West Darfur 6626203 44.23 15.37 7.26 0.33 0.39 0.48 0.55 97,936.00 43,320.21 West Darfur 6626204 49.01 18.39 9.18 0.83 1.08 1.43 1.73 245,568.00 120,348.92 West Darfur 6626205 68.17 29.66 16.29 0.31 0.56 0.86 1.15 91,438.00 62,333.05 West Darfur 6626206 58.19 22.88 11.75 0.51 0.79 1.10 1.36 150,947.00 87,842.02 South Darfur 6636302 46.69 17.23 8.51 1.44 1.80 2.35 2.80 428,808.00 200,195.64 South Darfur 6636303 60.47 24.56 12.89 0.04 0.07 0.09 0.12 11,994.00 7,253.17 South Darfur 6636304 56.08 21.72 11.03 0.53 0.79 1.08 1.33 156,341.00 87,681.48 South Darfur 6636305 54.33 20.50 10.21 2.39 3.46 4.62 5.58 709,985.00 385,747.55 South Darfur 6636306 56.13 21.62 10.95 0.82 1.23 1.67 2.05 243,551.00 136,716.84 South Darfur 6636309 34.25 10.73 4.70 1.35 1.23 1.37 1.45 400,784.00 137,286.68 South Darfur 6636310 64.62 26.82 14.24 0.92 1.59 2.34 3.01 274,755.00 177,553.03 South Darfur 6636311 40.97 13.55 6.17 1.06 1.16 1.36 1.49 314,900.00 129,013.89 Central Darfur 6646407 62.94 25.94 13.71 0.40 0.67 0.98 1.26 119,475.00 75,201.97 Central Darfur 6646408 66.31 28.20 15.25 0.11 0.19 0.29 0.38 32,245.00 21,380.81 Central Darfur 6646409 67.68 29.06 15.78 0.04 0.07 0.10 0.14 11,265.00 7,624.09 Central Darfur 6646411 57.56 22.36 11.37 0.59 0.91 1.25 1.54 175,870.00 101,227.21 58 Level Severity Contribution of Poverty Head Poverty Population Population Number of Locality of Head Poverty Severity of Region State Count Gap Share Size Poor Code Poverty Count Gap Poverty Central Darfur 6646412 76.58 36.29 21.06 0.11 0.22 0.37 0.52 32,119.00 24,597.46 Central Darfur 6646413 48.18 17.16 8.25 0.09 0.12 0.15 0.17 27,540.00 13,268.22 East Darfur 6656501 50.92 18.96 9.40 0.61 0.83 1.10 1.32 182,629.00 92,998.18 East Darfur 6656507 50.49 18.45 9.00 1.51 2.03 2.63 3.10 447,965.00 226,199.43 East Darfur 6656508 36.33 11.49 5.07 0.68 0.66 0.74 0.79 202,180.00 73,448.15 East Darfur 6656512 41.87 14.14 6.55 0.76 0.85 1.01 1.14 225,664.00 94,477.75 Source: Authors’ calculations based on the 2014/15 NHBPS and the 2008 Population and Housing Census. 59 Appendix F: Census Non-monetary Indicators by Administrative Units Table F.1: Population Characteristics by Region and State Sex Population Structure Literacy Education Level Not Adult Youth 0–14 15–64 65+ Applicabl Male Female (15+ (15–24 None Primary Secondary Tertiary Years Years Years e/Do Not Years) Years) Know SUDAN 49.57 50.43 42.57 53.78 3.65 32.61 13.48 4.54 40.16 7.97 4.09 43.25 49.18 50.82 36.29 58.11 5.60 44.97 16.43 10.02 45.28 11.91 4.93 27.86 Northern Northern 48.88 51.12 34.51 59.56 5.93 47.62 17.29 8.98 47.23 13.54 5.24 25.01 River Nile 49.35 50.65 37.33 57.27 5.40 43.43 15.92 10.64 44.11 10.93 4.74 29.58 52.12 47.88 42.05 54.73 3.22 26.52 10.98 3.47 36.64 4.87 1.81 53.20 Red Sea 55.86 44.14 37.98 59.50 2.52 26.16 10.01 4.41 29.61 6.19 2.26 57.53 Eastern Kassala 51.55 48.45 42.07 54.60 3.32 24.24 10.38 2.32 35.53 4.01 1.51 56.65 Al-Gedarif 48.99 51.01 46.23 49.97 3.80 29.81 12.76 3.97 45.93 4.56 1.73 43.82 Khartoum Khartoum 51.67 48.33 34.91 61.76 3.33 52.69 20.43 4.20 46.93 17.39 11.67 19.82 47.58 52.42 42.28 53.47 4.26 35.66 14.98 5.71 45.28 8.16 3.41 37.44 Al-Gezira 47.34 52.66 40.58 55.05 4.37 41.41 16.90 5.83 49.05 10.70 4.58 29.84 Central White Nile 47.60 52.40 42.44 53.26 4.30 32.77 14.27 4.75 43.69 7.29 3.12 41.15 Sinnar 47.37 52.63 43.22 52.48 4.31 32.63 13.88 7.02 44.15 5.65 1.94 41.24 Blue Nile 48.90 51.10 47.66 48.71 3.63 22.00 10.03 5.21 33.67 2.68 1.13 57.30 47.50 52.50 46.92 49.18 3.89 21.98 9.93 5.24 33.73 3.96 1.54 55.53 North Kordofan 47.23 52.77 45.71 50.14 4.15 21.91 10.01 6.69 32.70 4.06 1.56 54.99 Kordofan South Kordofan 47.95 52.05 47.85 48.44 3.71 24.69 10.66 3.50 38.70 4.50 2.22 51.07 West Kordofan 47.65 52.35 48.23 48.16 3.60 20.30 9.32 4.05 32.08 3.44 1.04 59.39 49.83 50.17 47.93 49.19 2.88 21.41 9.65 2.36 34.02 3.56 1.50 58.56 North Darfur 50.08 49.92 47.53 49.35 3.12 25.20 11.89 3.09 39.60 5.02 2.26 50.02 West Darfur 47.50 52.50 47.33 49.04 3.63 23.11 10.13 1.36 38.64 2.87 1.31 55.83 Darfur South Darfur 50.44 49.56 48.48 48.85 2.67 20.47 9.00 2.48 32.14 3.32 1.32 60.76 Central Darfur 46.94 53.06 49.99 46.99 3.02 11.14 4.65 1.30 21.87 0.51 0.20 76.12 East Darfur 50.73 49.27 47.07 50.65 2.29 18.66 8.26 1.75 28.31 2.86 1.00 66.08 60 Table F.2: Households Characteristics by Region and State Sanitation Water Own House Toilet Shared Bucket/No Toilet Improved* With Network Own Rent Free Yes No Yes No Yes No Yes No SUDAN 54.66 45.34 43.11 56.89 74.01 25.99 37.06 62.94 87.30 7.90 4.79 31.06 68.94 20.19 79.81 79.05 20.95 67.58 32.42 89.51 6.14 4.35 Northern Northern 19.52 80.48 14.16 85.84 92.22 7.78 78.23 21.77 88.01 5.72 6.27 River Nile 37.77 62.23 23.69 76.31 71.40 28.60 61.40 38.60 90.39 6.38 3.23 66.80 33.20 61.48 38.52 55.07 44.93 26.80 73.20 90.80 5.88 3.32 Red Sea 72.15 27.85 68.41 31.59 58.58 41.42 18.71 81.29 88.22 8.32 3.46 Eastern Kassala 65.65 34.35 60.65 39.35 55.57 44.43 29.44 70.56 91.74 5.26 3.01 Al-Gedarif 62.71 37.29 55.37 44.63 50.81 49.19 31.80 68.20 92.27 4.15 3.58 Khartoum Khartoum 24.54 75.46 7.76 92.24 84.82 15.18 78.77 21.23 69.35 22.30 8.34 58.73 41.27 45.79 54.21 69.54 30.46 50.05 49.95 89.49 4.73 5.78 Al-Gezira 56.55 43.45 43.12 56.88 80.92 19.08 72.03 27.97 87.25 5.23 7.52 Central White Nile 66.56 33.44 50.66 49.34 54.93 45.07 31.68 68.32 89.09 5.65 5.26 Sinnar 60.52 39.48 47.82 52.18 73.69 26.31 40.39 59.61 93.11 3.04 3.85 Blue Nile 48.84 51.16 43.80 56.20 45.79 54.21 10.29 89.71 94.29 3.26 2.45 62.46 37.54 51.87 48.13 69.84 30.16 9.94 90.06 92.99 3.76 3.25 North Kordofan 69.34 30.66 61.50 38.50 73.77 26.23 16.16 83.84 93.55 3.58 2.87 Kordofan South Kordofan 63.99 36.01 57.95 42.05 77.98 22.02 3.60 96.40 91.65 4.01 4.34 West Kordofan 50.55 49.45 32.57 67.43 58.20 41.80 4.33 95.67 92.99 3.89 3.12 66.64 33.36 55.83 44.17 83.79 16.21 7.20 92.80 92.35 4.60 3.05 North Darfur 55.61 44.39 48.97 51.03 83.14 16.86 4.47 95.53 94.49 3.92 1.59 West Darfur 58.94 41.06 47.32 52.68 87.51 12.49 11.57 88.43 89.26 6.42 4.32 Darfur South Darfur 73.22 26.78 58.81 41.19 81.86 18.14 8.73 91.27 90.90 5.56 3.54 Central Darfur 90.32 9.68 85.81 14.19 90.26 9.74 2.23 97.77 96.59 1.02 2.39 East Darfur 69.76 30.24 57.44 42.56 84.54 15.46 7.61 92.39 92.25 3.64 4.11 Source: Authors’ calculations based on the 2014/15 NHBPS and the 2008 Population and Housing Census. Note: * Improved water sources refer to water filtering, boreholes, hand pump, sand filter, and dug well but not in modalities. 61