Mongolia - HSES 2009

The 2009 HSES used the sampling frame which was devel­oped by the NSO based on 2005 population figures obtained from local registration offices. This updated sampling frame was of great importance because the spatial dis­tribution of the population had changed dramati­cally over the last years and any frame based on the Census 2000 would not be relevant anymore.
The design of the survey recognizes three ex­plicit strata: Ulaanbaatar, aimag centers, and soum centers and the countryside. In addition, the sample was implicitly allocated by districts and khoroos in Ulaanbaatar, and by aimags in rural areas. Each aimag center was an explicit sub-stratum. The selection strategy was different in each stratum: a two-stage process in urban areas and a three-stage process in rural areas. In Ulaanbaatar, 360 khesegs were initially selected, from each of which 10 households were chosen. In aimag centers, 12 or 24 bags were initially selected, and then 10 house­holds from each bag. In rural areas, first 52 soums, then 12 bags in each soum and fi­nally 8 households in each bag were selected. All 1,248 primary sampling units or clusters (units, bags or soums) were selected with a probability proportional to their sizes and were randomly allocated into twelve months of survey fieldwork.
The use of this sampling procedure means that households living in different areas of the coun­try have been selected with different probabilities. Therefore, in order to obtain representative sta­tistics for each stratum and for the coun­try as a whole, it was necessary to use sampling weights. The weight which was assigned to each household corresponds to the inverse of the selection probability and takes the sampling strategy into account.
The sample of 11,232 households was allocated as follows: 3,600 in Ulaanbaatar, 2,640 in aimag centers and 4,992 in rural areas and soum centers. However, the actual sample size used for this analysis is slightly smaller: 3,572 households in Ulaanbaatar; 2,639 in aimag centers; and 4,987 in rural areas and small towns. The difference is explained by 60 households, for which complete information was unavailable and were thus, excluded.