Sampling Procedure
The sample selection stratifies the targeted slums by flood proneness and the level of poverty (Erman et al., 2018) as the following:
1. Slum areas were identified by combining the definition for informal settlement used by Accra Metropolitan Assembly (AMA) and UN Habitat (2011) and a slum index score developed by Engstrom et al. (2017). Enumeration areas (EAs) were added to the sample frame if they were defined as being in a slum area using the following definition: i) they were fully inside the areas defined as informal settlement according to AMA and UN Habitat’s definition and ii) had a slum index value higher than 0.7.
2. Enumeration areas in the sample frame were categorized as low poverty and high poverty by using a neighborhood-level poverty estimate created by Engstrom et al. (2017).
3. Enumeration areas in the sample frame were also categorized as flood-prone and not flood-prone using average elevation levels in the enumeration area. High flood risk areas are defined as below 17.5 meters (based on average elevation of areas flooded in the 2015 flood) and low risk areas as above 35 meters (the elevation level, above which there were no reported flooding during the 2015 flood).
4. Four neighborhoods in which all EAs were considered high risk and 4 neighborhoods in which all EAs were considered low-risk and one neighborhood with a mix of high and low-risk EAs were selected for the sample frame. In all selected neighborhoods, all EAs were defined as slum areas. The neighborhoods selected were Korle Lagoon Area, Jamestown, Gbegbeyise and Korle Dudor as high flood risk areas, and Abeka, Accra New Town, Mamobi, and Nima as low flood risk areas and Pig Farm, which includes both high and low flood risk areas. Neighborhoods are indicated in Figure 1 in a map of Accra. This administrative division was extracted from Engstrom et al. (2013).
5. The EAs in the selected neighborhoods were stratified into four categories: i) high flood risk and high poverty incidence; ii) low flood risk and high poverty incidence; iii) high flood risk and low poverty incidence; iv) low flood risk and low poverty incidence, of all selected neighborhoods.
6. Two-stage sampling was applied; 12 EAs per strata were selected using Probability Proportion to Size (PPS) and then 20 households per selected EA were selected using random sampling after listing. The sample size was determined using power calculations.
The shapefile of the Accra neighborhoods can be found in the folder DPHS_AccraGhana_Neighbourhoods, among the resources made available. The neighborhood shapefile can be matched with the surveyed neighborhoods in the DPHS dataset (DPHS_AccraGhana_Data) through the key variable neighbourhood_code.
Reference list:
ENGSTROM, R., OFIESH, C., RAIN, D., JEWELL, H., AND WEEKS, J. (2013): “Defining neighborhood boundaries for urban health research in developing countries: A case study of Accra, Ghana”, Journal of Maps, 9(1), 36-42.
ENGSTROM, R., D., PAVELESKU, T., TANAKA, A., AND WAMBILE (2017): “Monetary and non-monetary poverty in urban slums in Accra: Combining geospatial data and machine learning to study urban poverty,” Work in Progress, The World Bank.
ERMAN, A., MOTTE, E., GOYAL, R., ASARE, A., TAKAMATSU, S., CHEN, X., MALGIOGLIO, S., SKINNER, A., YOSHIDA, N., AND HALLEGATTE, S. (2018): “The road to recovery: the role of poverty in the exposure, vulnerability and resilience to floods in Accra,” Policy Research Working Paper; No. 8469. World Bank, Washington, DC.