1. Sampling and Methodology
Stratification
The sampling frame was divided between economic zones (South Central, Puntland, and Somaliland), then further broken down between urban areas, rural areas and IDP camps.
Sampling
The survey sampling frame was constructed based on the UNFPA Population Estimation Survey of Somalia (PESS) and the first round of the World Bank Somalia High Frequency Survey (SHFS).
The sampling frame is not representative of the entire Somali population: i) nomads were not surveyed; ii) only accessible areas (on security reasons) were included. The statistics only account for the areas included in the sampling frame. As nomadic populations and populations living in areas assessed as inaccessible are likely to have lower levels of access to mobile money, the estimates are likely to overestimate penetration and volumes of mobile money, would one like to infer statistics on the entire Somali population.
The design was a stratified four-stage cluster sample, with Probability Proportional to Size (PPS) of selection for the districts and enumeration areas (EAs). Households within the EA were selected using a random walk method. Respondents within the household were selected through a randomization formula within the encoded questionnaire, among the adults above 16 present at the time of the interview.
The sample size was computed to achieve representative results per strata, with a margin of error of +/-3% and a confidence level of 95%.
The research team decided to oversample. In each EA, instead of interviewing 10 households, as per the sample size calculations, 12 households were interviewed. The rationale was to minimize the need to go back to an EA to redo invalid interviews, for security reasons.
Questionnaire
A questionnaire translated in Somali with over 110 questions was used.
Calculations and tools
The databased is composed of 1,796 valid and successful interviews. The research team used post-weighting adjustments to balance the sample along key demographics (to correct for some existing within-household coverage bias) and correct for the oversampling.
Other research modules
The quantitative data collection was complemented by a desk review, Focus Group Discussions with various categories to dive deeper into the target populations’ needs and current behaviors, In-Depth Interviews of sector experts, incl. government authorities and academics, Mobile Network Operators, remittance services providers, banks, etc. and visits to mobile money agents.
2. Vulnerability Index
A vulnerability index was created to provide an overview of the living conditions of the respondents and to measure their vulnerability. It is based on five different variables (housing, drinking water, cooking, toilets and roof). Each variable lists a range of modalities that are each categorized as either -1 ‘poor’, 0 ‘neutral’ or 1 ‘wealthy’. The points are tallied to provide a final score. The final index scales from -3 (most vulnerable) to 3 (least vulnerable). As per this index, 30% of the population can be considered as vulnerable or very vulnerable.