Multiple Indicator Cluster Survey 2003

Sampling Frame
Availability of a sampling frame was a major constraint in planning the survey. There were two sources of data as sampling frame, one is the Population Census-1979 data from Central Statistics Office (CSO) and the other was the National Immunization Day (NID) coverage data from the WHO and UNICEF. Both the sources have their own limitations and advantages. Census 1979 data are outdated with all the displacement due to war and fertility and mortality. However, the data were found organized for use in any purpose. On the other hand, the NID data were produced based on the NID coverage in every round, which could be considered as the most recent. But the areas with incomplete coverage will have faulty data because the number beyond the coverage is not known accurately. The format of the data in NID was also not standardized for all the provinces. Standardization would need sometime to make it user-friendly. A comparative analysis was carried out to address the sampling frame issue and it was apparent that Census data could be used for the purpose at the moment. A sampling design and the related survey methodology was henceforth developed for MICS in Afghanistan in 2003.

Sample Design
A two-stage cluster sampling strategy was decided for the survey where villages were treated as the primary sampling unit (PSU). The number of PSUs selected from each province is decided as 20. This again translates to a number of 27.5 households per PSU or cluster. Clusters were defined as the compact area segment of 55 households in each selected PSU.
The six major cities were treated as separate domain of the study. Hence they were separated out from the respective provinces. For each of these cities, except Kabul, 20 clusters are selected while the number is 25 for Kabul city. Under consideration of huge population and possible heterogeneity, this number is proposed for Kabul to achieve wider/better coverage.

Selection of Households
Every selected village was partitioned into several equal segments of roughly 55 households each. Segments were numbered according to a prescribed fashion, starting from southeast corner and move anti-clockwise. And the segment corresponding to the random number of the selection was selected for survey. Within each segment every alternative households were interviewed starting randomly from the 1st or 2nd household.

Stratification and post-stratification

Stratification for the survey is decided to be at two levels

1. 32 provinces as both strata and study domain, where 6 provinces containing the major cities had only the rural areas
2. 6 major cities as strata and study domain
3. Post-stratification was done for the 26 provinces where urban and rural areas were not separated at the time of sampling. This post stratification is used to separate out the data of the provincial urban areas (defined as the district of provincial head quarter) from that of rural areas.

Hence the estimates are available for

1. All 32 provinces (urban +rural)
2. 6 rural provinces, 6 major cities
3. National Metro cities, other urban areas (26 provincial towns) and
4. National Urban, National Rural, National (Urban+Rural)