05 Dec Volume 1: Methodology and Assessment of Data Quality of the Somali People
There has been only one successful census in Somalia, carried out in 1975. The 1986 census remained incomplete, and its results were never published. The PESS was conducted to provide updated information on the Somali population for planning purposes, as well as humanitarian work. It is the first comprehensive population survey to be carried out since the 1975 census. To collect information, the PESS selected a representative sample of clusters through a one-stage stratified-cluster sample design. These clusters formed the primary sampling units (PSUs), which were referred to as enumeration areas (EAs) in urban areas, settlements in rural areas, camps for internally displaced persons (IDPs), and water points for nomadic populations. In each of these areas, except for the water points, the PSUs comprised 50 to 149 households.
The survey methodology entailed determining sample sizes for urban, rural and nomadic areas, and internally displaced person camps, allocating the sample to the 18 pre-war regions, which formed the first strata. This was followed by the selection of clusters for urban and rural areas, and IDP camps. For nomadic populations, water points were grouped by type, following which the water points were selected for enumeration. The estimated sample size was 2,535 PSUs, which was then adjusted to 2,735, comprising 868 EAs, 1,104 villages, 735 water points (for the nomadic population) and 28 camps for the IDPs.
Data was processed in a standard manner in statistical hubs in Hargeisa, Garowe and Mogadishu. The string of processes began with the receipt of the filled-in questionnaires, manual code checking, data entry and processing/exporting of data from Census and Survey Processing System (CSPro) into Statistical Package for the Social Sciences (SPSS), and finally the construction of weights, which were included in the data set.
The PESS team evaluated the extent of inaccuracies in age and sex reporting in the data by applying the single year age and sex distribution, Whipple’s Index, Myer’s Blended Index and UN Age Sex Accuracy Index. The assessment of the quality of age and sex data indicates an undercount in the age group 0-4 years, “age heaping” on ages with terminal digits “0” and “5”, and relatively low numbers of the male population in the ages 20-39 years, which may indicate
emigration of men in the working ages. Statistical methods of “smoothing” were applied to restructure the reported age and sex data to address age misreporting. Different methods/approaches are presented in this volume. Strong smoothing provided ideal estimates for the age and sex data of Somali people. The use of smoothed data on the macro-level is appropriate for national population projections and other specific applications of national planning. The smoothing showed that reported ages are relatively inaccurate at younger ages, and is more pronounced for males than females. However, for the calculation of complex demographic indicators and the breakdown of indicators by regions and type of residence, it is not possible to use smoothed data at this stage, as additional work and analysis would be required. The analytical volumes of the PESS include, therefore, only data that are not smoothed.
The computed standard errors indicate confidence in and high reliability of PESS estimates. The Coefficient of Variation (CV), a measure for reliability for most estimates is assessed to be good (CV <= 15%) based on a classification scheme developed for the American Community Survey (ACS) conducted by the United States Bureau of the census. The Coefficient of Variation for most estimates was less than 10 percent, which is below 15 percent, the upper margin for estimates assessed to be good, which generally indicates that the quality of the PESS meets international standards.